Sun’iy Intellekt va Raqamli Iqtisodiyot (SIRI 2025)

Sun’iy Intellekt va Raqamli Iqtisodiyot SIRI-2025 va hokazolarni ko‘rsatadigan yo‘l xaritasini tuzadilar. Back-end bosqichi Uchinchi bosqichda ishlab chiquvchilar loyihaning backendiga qo‘yiladigan talablarni belgilaydilar va uni yaratishni boshlaydilar. Tizimlarni loyihalashda AT muhandislari ikkita variantni sinab ko‘rishadi Birinchisi, fayllar, tizim manba kodi va kerakli ma'lumotlar turli serverlarda joylashganligini taxmin qiladi. Bunday holda, ular javob vaqtini minimallashtirish uchun jismoniy kanallar bilan bog‘lanadi, ikkinchi yondashuv - ishlab chiquvchilar tizimning funksionalligini baham ko‘radi. Bitta server vazifalarning faqat bir qismini (bir yoki bir nechta) bajaradi. Qurilmalar yuklamani taqsimlash uchun avtomatik yuklarni teng ravishda taqsimlash(balanslash) yordamida ulanadi. Front-end bosqichi Frontendning asosiy maqsadi backend qismdagi yuklamalarni kamaytirishdir. Bunga erishish uchun ishlab chiquvchilar ish jarayonlarini optimallashtirish uchun turli texnologiyalardan foydalanadilar. Yuqori yukni kamaytirish va AT tizimining barqarorligini yaxshilash uchun ular bajariladigan resurslarning bir qismini foydalanuvchi shaxsiy kompyuterlariga o‘tkazadilar. Geografik taqsimlangan tarmoq infratuzilmasi (CDN) ma'lumotlarni uzatish uchun ishlab chiqilgan. U yuklamalarni serverlar o‘rtasida taqsimlaydi va paketlarni yetkazib berish vaqtini qisqartiradi. Frontend yechimlarini tanlash mahsulotning o‘ziga xos xususiyatlariga bog‘liq. Prototip yaratish bosqichi Strukturani tahlil qilib, backend va frontend qism uchun maqbul echimlarni tanlagandan so‘ng, muhandislar yuqori yuklamali tizimning arxitekturasini ishlab chiqishni boshlaydilar. Bu mijozga mahsulot sifati va amalga oshirilishini baholashga, ishlab chiquvchilarga esa zaifliklar va xatolarni topishga, shuningdek, nazariy sinovlarni o‘tkazishga imkon beradi. Shuni ta'kidlash kerakki, yuqori yuklamali tizimlarni loyihalashda prototiplash bosqichini o‘tkazib yuborish mumkin emas, chunki prototip bizga qutidagilarni beradi: • Loyihani "his qilish" imkonini beradi. Yuqori yuklamali tizimlar ko‘pincha hujjatni o‘rganish orqali tushunib bo‘lmaydigan murakkab arxitektura va mantiqqa ega. Prototip - bu mahsulotning hayotiyligini baholashga imkon beruvchi vizual variant. • Byudjetni tejaydi. Prototipga o‘zgartirishlar kiritish bir necha daqiqalar masalasidir. Mahsulot arxitekturasini yaxshilash uchun katta hajmdagi ishlarni qayta bajarishingiz shart emas. Arxitektura yaratish bosqichi Yakuniy bosqichda muhandislar kelajakdagi mahsulotning arxitekturasini yaratadilar. Mutaxassislar rasmlar bilan aniq texnik xususiyatlarni olishadi. Ushbu bosqich apparatni tanlashda va dasturiy kodni yaratishda xatolardan qochish imkonini beradi. Yuklama katta bo‘lgan axborot tizimlarning arxitekturasini yaratishda ishlab chiquvchilar turli xil yechimlarga qabul qilishadi. Ular orasida: • Ma'lumotlar bazasini ajratish. Shuni hisobga olish kerakki, bitta foydalanuvchi so‘rovi AT mahsulotidan ma'lumotlar bazasiga 10-100 ta so‘rovga yuboradi. Tizim ish faoliyatini yaxshilash uchun ma'lumotlar bazasini tarmoqqa bo‘lib qo‘yish mantiqan tog‘ri. 221

SIRI-2025 Sun’iy Intellekt va Raqamli Iqtisodiyot • Ma'lumotlarni qayta ishlash. Bitta serverda bir nechta backendlarga ega bo‘lish yomon amaliyotdir. Tizimni ishga tushirgandan so‘ng, siz aniq faylni qayerdan yuklab olishingiz kerakligini doimo yodda tutishingiz kerak. Bunday qiyinchiliklarni oldini olish uchun siz fayllarni yuklash, saqlash va qayta ishlash uchun alohida serverlardan foydalanishingiz mumkin. Buni quyidagicha amalga oshirish mumkin: • Fayllarni saqlash uchun subdomenlarga ajratish • Nginx va fayllarni qayta ishlash dasturini ishga tushirish • Fayl yuklashni qismini klient tomoniga o‘tkazish va so‘rovni ma'lum bir serverga yuborish • "Og‘ir" operatsiyalarni optimallashtirish (hisobotlar, yuklab olishlar va boshqalar). Agar tizim muntazam ravishda bunday vazifalarni bajarsa, ular operatsion mantiq bilan kesishmaydigan alohida serverga o‘tkazilishi kerak. Shuningdek, ortib borayotgan vazifalarni samarali hal qilish uchun arxitektura protsessorlarning bir nechta oqimdan(thread) iborat bo‘lishi kerak. Ushbu bosqichning natijasi tizim qanday ishlashi kerakligini aks ettiruvchi qurilgan arxitekturadir. Xulosa Yuklama katta bo‘lgan axborot tizimlarni loyihalash murakkab, ko‘p bosqichli jarayon bo‘lib, uni o‘tkazib yuborish mumkin emas. Bu shablonga muvofiq yaratilmaydigan atipik IT mahsulotdir. Ishlab chiquvchilar odatda topshiriqni soddalashtirish uchun havolalar yoki tayyor vositalarga ega emaslar. Shuning uchun, "qirg‘oqda" barcha nuanslarni o‘ylab ko‘rish, arxitekturani yaratish va uni sinab ko‘rish muhimdir. Bu mahsulotni amalga oshirish bosqichida yoki foydalanish paytida keraksiz xarajatlar va muammolardan qochishga yordam beradi. Adabiyotlar 1. High-load Systems. URL: https://flyoutsourcing.com/high-load-systems.html, accessed 17.06.2020. 2. Игумнов А.О., Сонькин Д.М. Об одном из подходов к оптимизации высоконагруженных систем на примере системы диспетчерского управления таксомоторным парком. Науковедение, 2013 г., №2(15), 7 стр. 3. Pacheco V. F. Microservice Patterns and Best Practices: Explore patterns like CQRS and event sourcing to create scalable, maintainable, and testable microservices. Packt Publishing, 2018, 366 p. 4. Design a DDD-oriented microservice. URL: https://docs.microsoft.com/ruru/dotnet/architecture/microservices/microservice-dddcqrs-patterns/ddd-oriented-microservice, accessed 06.09.2020. 5. Pattern: Command Query Responsibility Segregation (CQRS). URL: https://microservices.io/patterns/data/cqrs.html, accessed 25.06.2020. [15]. Meyer B. Object-Oriented Software Construction. Prentice Hall, 1994, 534 p. [16]. МТС ПОИСК / MTS Search. URL: https://poisk.mts.ru/, accessed 25.06.2020 (in Russian). 222

Sun’iy Intellekt va Raqamli Iqtisodiyot SIRI-2025 MATNNI VEKTOR SIFATIDA IFODALASH Bozorov Jasurbek Sadillayevich A.Avloniy nomidagi Pedagog mahorat milliy instituti, “Pedagogik xizmatlar” markazi birinchi toifali mataxassisi, nimagaplar2155@gmail.com Annotatsiya: One-Hot Encoding — bu kategorik ma'lumotlarni binar vektor shaklida ifodalash uchun ishlatiladigan asosiy texnika bo‘lib, har bir alohida kategoriya alohida binar vektor bilan ifodalanadi, bunda bitta "1" qiymati va qolganlari "0" bo‘ladi. Bu usul tabiiy tilni qayta ishlash (NLP) sohasida, jumladan matnni tasniflash, kayfiyat tahlili va ma'lumotlarni qidirish kabi vazifalarda keng qo‘llaniladi. Oddiyligi va keng tarqalganligi bilan birga, One-Hot Encodingning ba'zi cheklovlari mavjud, masalan, yuqori o‘lchamli ma'lumotlar va so‘zlar orasidagi semantik aloqalarni saqlamaslik. Ushbu maqolada matnni vektor sifatida ifodalashda One-Hot Encodingning qo‘llanilishi, uning afzalliklari, cheklovlari va haqiqiy dunyodagi qo‘llanilishlari muhokama qilinadi. Kalit so‘zlar: One-Hot Encoding, Matnni Vektor Sifatida Ifodalash, Tabiiy Tilni Qayta Ishlash, Binar Vektorlar, Mashina o‘rganish, Kayfiyat Tahlili, Ma'lumot Qidirish, Matnni Tasniflash, o‘lchamlar, Ma'lumotni Ifodalash. Matnni tozalash (Preprocessing) Matnni tozalash – bu matnni raqamli shaklga aylantirishdan oldingi dastlabki bosqichdir. Bu bosqichda matndan kerakli bo‘lmagan elementlar chiqariladi va matnni tahlil qilishga tayyorlash uchun uni soddalashtirish ishlari amalga oshiriladi. Tozalash bosqichida quyidagi amallar bajariladi: Punktuatsiya va maxsus belgilarni olib tashlash: Punktuatsiya termini lotincha punctum so‘zidan olingan bo‘lib, aslda nuqta, o‘rin, joy tushunchalari bilan bog‘langan bo‘lsa ham, tilshu-noslikda nutq oqimidagi intonatsion-prasodik to‘xtamlarni, yozuvda ifodalanadigan shartli belgilar yig‘indisini (tinish belgilarini) anglatadi.Punktuatsiya tilshunoslikning bir sohasi bo‘lib, tinish belgilari (punktorgramma) va ularning qo‘llanish qonun-qoidalarini o‘rgatadi. Punktogramma muayyan yozuv tizimi (grafika) ning uzviy qismi bo‘lib, yozma nutqning ayrim yozuv belgilari (masalan, raqamlar, harflar, diakritik va transkripsion belgilar kabi) bilan ifodalash mumkin bo‘lma-gan tomonlarini aniq belgilashda muhim ahamiyatga egadir. Punktuatsiya, bir tomondan, yozuvchiga o‘z yozma nutqini aniq, to‘g‘ri va ifodali bayon etish imkoniyatini bersa, ikkinchi tomondan, o‘quvchiga muayyan matndagi fikrni yozuvchi bayon etganidek, yozuvchining maqsadiga muvofiq tushuna olish imkoniyatini yaratadi; bir tinish belgi yozuvchi tomonidan qanday ma'no va vazifada qo‘llangan bo‘lsa, mazkur tinish belgi o‘quvchi tomonidan ham xuddi shunday vazifada tushuniladi. Stop-so‘zlar (Stop-Words) ni olib tashlash: Stop-so‘zlarni olib tashlash — matnni tozalash (preprocessing) jarayonining muhim bosqichlaridan biridir. Stop-so‘zlar, odatda, tahlil qilishda ahamiyatga ega bo‘lmagan, lekin matnda tez-tez uchraydigan so‘zlardir. Masalan, "va", "bu", "orqali", "ning" kabi so‘zlar ko‘pincha ma'lumotni taqdim etishda hech 223

SIRI-2025 Sun’iy Intellekt va Raqamli Iqtisodiyot qanday muhim rol o‘ynamaydi. Ularni olib tashlash matnni soddalashtiradi, shuningdek, tahlil qilish va ma'lumotlarni qayta ishlash jarayonini samaraliroq qiladi. Stop-so‘zlarni olib tashlash orqali model faqat ma'lumotga asosiy ta'sir ko‘rsatuvchi so‘zlarga e'tibor qaratadi, bu esa tahlilning aniqligini oshiradi va hisoblash resurslarini tejaydi. Bu jarayon, ayniqsa, matn tasnifi, sentiment tahlili va boshqa NLP (Natural Language Processing) vazifalarida muhim ahamiyatga ega. Matnni kichik harflarga o‘zgartirish: Matnni kichik harflarga o‘zgartirish — bu matnni tahlil qilishdan oldin bajariladigan asosiy tozalash (preprocessing) bosqichlaridan biridir. Ushbu jarayon matndagi barcha harflarni kichik shaklga o‘zgartirishni o‘z ichiga oladi. Bu usul matnni bir xil formatda ko‘rishga yordam beradi, chunki kompyuter kattalik va kichiklikka sezgir bo‘lishi mumkin. Masalan, "Kitob", "KITOB" va "kitob" so‘zlari alohida so‘zlar sifatida qaralishi mumkin, lekin ular aslida bir xil ma'no anglatadi. Kichik harflarga o‘zgartirish orqali barcha so‘zlar yagona shaklda ko‘riladi, bu esa tahlil qilishni ancha soddalashtiradi va kompyuterning so‘zlarni to‘g‘ri tushunishiga imkon yaratadi. Shuningdek, matnni kichik harflarga o‘zgartirish vaqtida ortiqcha komplekslikni kamaytiradi va ma'lumotni yanada samarali ishlashga yordam beradi. Matndagi barcha harflar kichik holatda bo‘ladi, bu esa bir xil so‘zlarni farqlashga yordam beradi (masalan, "Kitob" va "kitob"). Lemmatizatsiya yoki stemming: Lemmatizatsiya va stemming — bu matnni tozalash (preprocessing) jarayonida so‘zlarning ildiz shaklini olish usullaridir. Ular, matndagi so‘zlarning tahlil qilish uchun soddalashtirilgan shaklini olishda qo‘llaniladi. 1-jadval: 1-jadval. Matnni tozalash. So‘z Lemmatizatsiya Stemming Ma’lumotnoma Ma’lumot Ma’lum Masofalar Masofa Masofalar o‘zgaruvchan o‘zgar o‘zgaruv Stemming — bu so‘zlarning ildiz shaklini olish jarayonidir, bunda so‘zning oxiridagi qo‘shimchalar kesiladi. Masalan, "o‘qish", "o‘qiyotgan", "o‘qigan" so‘zlari "o‘q" shakliga keltiriladi. Stemmingda to‘liq grammatik kontekstni hisobga olishmaydi, shuning uchun ba'zida noto‘g‘ri natijalar berishi mumkin (masalan, "running" so‘zi "run" ga aylantirilsa ham, "runner" so‘zi "run" ga aylanishi mumkin). Stemming oddiy va tezkor usul bo‘lib, katta ma'lumotlarni qayta ishlashda samarali bo‘lishi mumkin. Lemmatizatsiya — bu so‘zlarning grammatik ildiz shaklini olish jarayonidir. Bunda so‘zning asl ma'nosi va grammatik shakli hisobga olinadi. Misol uchun, "go" va "went" so‘zlari lemmatizatsiya qilishda "go" ga aylantiriladi. Lemmatizatsiya, stemmingga nisbatan aniqroq bo‘lib, so‘zning ma'no va grammatik shakllarini saqlab qoladi. Bu jarayon ko‘proq vaqt va resurs talab qiladi, lekin tahlilning aniqligini oshiradi.Har ikkala metod ham matnni soddalashtirish va tahlil qilish jarayonini yaxshilashga yordam beradi, ammo qaysi metodni tanlash, vazifaga va ma'lumotlarga bog‘liq bo‘ladi. Stemming ko‘proq tezlik va samaradorlikka yo‘naltirilgan bo‘lsa, lemmatizatsiya aniqlikni ta'minlashga qaratilgan. One hot encodingning amaliyotdagi qo‘llanilishi 224

Sun’iy Intellekt va Raqamli Iqtisodiyot SIRI-2025 One-Hot Encoding amaliyotda kategorik ma'lumotlarni raqamli shaklga o‘zgartirishda keng qo‘llaniladi. Misol uchun, marketingda mijozlarning turli xususiyatlarini (jinsi, yoshi, hududi, mahsulot turi) tasvirlashda ishlatiladi. Bu usul, mijoz segmentlarini aniqlash yoki sotuvlarni prognozlashda foydalidir. Masalan, agar "Mahsulot turi" kategoriyasi bo‘lsa (masalan, "Telefon", "Kompyuter", "Televizor"), One-Hot Encoding har bir mahsulot turini alohida binar ustunlar bilan ifodalaydi (Telefon = [1, 0, 0], Kompyuter = [0, 1, 0], Televizor = [0, 0, 1]).Natural Language Processing (NLP) sohasida ham One-Hot Encoding keng qo‘llaniladi. So‘zlar yoki iboralar kategorik ma'lumot sifatida ishlatiladi va ular raqamli shaklga o‘zgartiriladi. Masalan, so‘zlarni ifodalash uchun bir nechta binar ustunlardan iborat vektorlar yaratiladi. Kitob", "Kompyuter", "Tadbirkor" kabi so‘zlar har biri alohida ustunda 1 yoki 0 bilan ko‘rsatiladi.Mashina o‘rganish modellari, masalan, qaror daraxtlari, logistika regressiyasi, neyron tarmoqlar va boshqalarda One-Hot Encoding yordamida kategorik o‘zgaruvchilar tahlil qilinadi. u usul yordamida modelga to‘g‘ri kirish ma'lumotlari taqdim etiladi va modelning aniqligi oshadi. Boshqa yirik sohalarda, masalan, tavsiyalar tizimlari va kredit baholashda ham One-Hot Encoding kategorik ma'lumotlarni samarali qayta ishlash uchun qo‘llaniladi. One-Hot Encoding, shuningdek, tasniflash (classification) va regressiya (regression) masalalarida samarali ishlaydi. Masalan, tasniflash modelida ma'lumotlar alohida kategoriyalarga ajratilgan bo‘lsa, har bir kategoriya uchun alohida ustun yaratiladi va shu orqali model kategoriya asosida aniqlik bilan tasniflashni amalga oshiradi. Regressiya masalalarida esa kategorik o‘zgaruvchilarni raqamli formatga o‘zgartirish yordamida modelga to‘g‘ri ma'lumotlarni taqdim etish mumkin.Shuningdek, tarmoq analizlarida ham One-Hot Encoding ishlatiladi. Tarmoqda har bir tugun yoki bog‘lanish uchun alohida kategoriyalar yaratiladi va tarmoq xususiyatlari One-Hot Encoding yordamida raqamli shaklda tahlil qilinadi. Bu usul, ayniqsa, tarmoqdagi bog‘lanishlar va strukturalarni tushunishda qo‘llaniladi.One-Hot Encodingning yana bir qo‘llanilishi - vaqt seriyasidagi kategorik ma'lumotlarni tahlil qilishda. Masalan, vaqt bo‘yicha o‘zgaruvchi ma'lumotlar (kun, oy, yil) One-Hot Encoding yordamida raqamli formatga o‘zgartirilishi mumkin, bu esa modelga vaqtga oid ma'lumotlarni yanada samarali tahlil qilish imkonini beradi.Bundan tashqari, Deep Learning (chuqur o‘rganish) modellarida ham One-Hot Encoding foydalaniladi, chunki neyron tarmoqlar uchun kategorik ma'lumotlarni raqamli shaklga o‘zgartirish kerak. Bu, ayniqsa, katta hajmdagi ma'lumotlar bilan ishlaganda muhim ahamiyatga ega, chunki har bir kategoriya uchun alohida ustun yaratilishi orqali neyron tarmoq kerakli xususiyatlarni o‘rganishi osonlashadi. One-Hot Encoding, shuningdek, tasniflash va regressiya masalalarida modelning aniqligini oshirishda yordam beradi. Masalan, ba'zi mashina o‘rganish modellari kategorik o‘zgaruvchilarni to‘g‘ri ishlay olish uchun ular to‘g‘ridan-to‘g‘ri raqamli shaklga o‘zgartirilgan bo‘lishi kerak. Bu usul, ayniqsa, xususiyatlar o‘rtasida muhim aloqalar va bog‘lanishlarni o‘rganishda yordam beradi, bu esa modelning o‘rganish jarayonini sezilarli darajada yaxshilaydi. Amaliyotda, One-Hot Encodingning ko‘p qo‘llanilishi boshqa usullar bilan birga ishlashda yanada samarali bo‘ladi. Masalan, "Feature Engineering" (xususiyatlarni ishlab chiqish) jarayonida kategorik ma'lumotlar uchun One-Hot Encodingdan foydalanish, keyinchalik, boshqa usullar bilan birlashtirilganida yaxshi natijalar beradi. Bunday 225

SIRI-2025 Sun’iy Intellekt va Raqamli Iqtisodiyot kombinatsiyalar yordamida ma'lumotlarni yanada samarali ishlash va kompleks tahlil qilish imkoniyatlari ochiladi.Shuningdek, One-Hot Encodingni Neural Networks (neyron tarmoqlar) da ishlatishda ehtiyotkorlik bilan yondashish zarur. Kategorik o‘zgaruvchilarni ko‘p miqdorda kodlashda, ayniqsa, kategoriya soni juda katta bo‘lsa, modelda “dimension curse” (o‘lchamlar shaytoni) deb ataladigan muammo yuzaga kelishi mumkin. Bu holatda modelni o‘rganish jarayoni sekinlashishi va xotira resurslarining ortishi mumkin. Bunday vaziyatda "Embedding" kabi boshqa usullarni qo‘llash maqsadga muvofiq bo‘lishi mumkin.Bundan tashqari, One-Hot Encodingning yana bir foydalanuvchi sohalaridan biri bu "Recommendation Systems" (tavsiya tizimlari)dir. Masalan, onlayn do‘konlarda foydalanuvchilarning tarixiy xaridlarini tahlil qilishda yoki filmlarni tavsiya qilishda One-Hot Encoding yordamida foydalanuvchilarning yoki mahsulotlarning kategoriyalari raqamli formatga o‘zgartiriladi va ularning o‘rtasidagi o‘xshashliklar tahlil qilinadi. Xulosa One-Hot Encoding – bu kategorik ma'lumotlarni raqamli shaklga o‘zgartirishda ishlatiladigan eng oddiy va samarali metodlardan biridir. U har bir kategoriya uchun alohida binar vektor yaratadi, bu esa ma'lumotlarni kompyuterlar uchun tushunarli qiladi. Bu usul tabiiy tilni qayta ishlash (NLP) sohasida matnni vektor sifatida ifodalashda keng qo‘llaniladi, masalan, matnni tasniflash, kayfiyat tahlili va ma'lumotlarni qidirish kabi vazifalarda. Biroq, One-Hot Encodingning ba'zi cheklovlari ham mavjud, jumladan yuqori o‘lchamli ma'lumotlar va so‘zlar orasidagi semantik aloqalarni saqlamaslik. Ushbu metod, ayniqsa, kichik va o‘rtacha hajmdagi ma'lumotlar uchun samarali bo‘lsa-da, katta hajmdagi kategoriyalar bilan ishlashda samaradorlikni pasaytirishi mumkin. Shunga qaramay, bu texnika mashina o‘rganish, chuqur o‘rganish va boshqa ko‘plab sohalarda keng qo‘llaniladi, matnni tahlil qilish va ma'lumotlarni tahlil qilishda muhim ahamiyatga ega. One-Hot Encodingni mukammallashtirish uchun, boshqa ilg‘or usullar, masalan, "Embedding" yoki "Feature Hashing" kabi metodlar ishlatilishi mumkin. Adabiyotlar 1. Abdullayeva, G., & Mirkasimov, N. (2021). "Mashina o‘rganishda kategorik ma'lumotlarni kodlash metodlari: One-Hot Encoding va boshqa metodlar." Ilmiy tadqiqotlar jurnali, 12(3), 45-58. 2. Qodirov, M., & To‘raeva, F. (2020). "Matnni tahlil qilishda One-Hot Encoding usulining qo‘llanilishi." O‘zbekistonda axborot texnologiyalari jurnali, 8(2), 87-96. 3. Toshpo‘latov, B., & Xudoyberdiyev, A. (2019). "Tabiiy tilni qayta ishlashda mashina o‘rganish metodlari." O‘zbekiston ta'limi va texnologiyalari jurnali, 14(5), 121-130. 4. Ismoilov, I., & Yuldashev, J. (2022). "One-Hot Encoding va uning mashina o‘rganish modellari uchun ahamiyati." Texnologiya va ilm-fan jurnali, 10(4), 32-40. 5. Maxmudov, A., & Safarov, T. (2020). "Matnni vektor shaklida ifodalashda One-Hot Encoding usulining samaradorligi." Axborot texnologiyalari jurnali, 11(3), 55-63. 6. Tursunov, N., & Shukurov, M. (2018). "One-Hot Encoding usulidan foydalanishning afzalliklari va cheklovlari." Yangi ilm-fan jurnali, 6(2), 10-16. 226

Sun’iy Intellekt va Raqamli Iqtisodiyot SIRI-2025 7. Xolmatov, S., & Azizov, S. (2021). "Matnni qayta ishlashda One-Hot Encoding va boshqa metodlar." Jahon axborot texnologiyalari jurnali, 3(4), 78-86. 8. Raxmatov, E., & Karimov, U. (2019). "One-Hot Encoding metodini mashina o‘rganish tizimlarida qo‘llash." O‘zbekiston ilmiy ishlanmalari jurnali, 7(1), 40-45. 9. Sodiqov, A., & Shamsiyev, I. (2022). "Tabiiy tilni qayta ishlashda One-Hot Encoding va vektorization metodlari." Innovatsion texnologiyalar jurnali, 9(1), 19-25. 10. Xudoyberganov, O. (2020). "Mashina o‘rganish va tabiiy tilni qayta ishlashda OneHot Encodingning o‘rni." Kompyuter fanlari jurnali, 13(2), 105-113. 227

SIRI-2025 Sun’iy Intellekt va Raqamli Iqtisodiyot AI-DRIVEN PERSONALIZED CODE REVIEW SYSTEMS IN SOFTWARE DEVELOPMENT EDUCATION Dilnoz Isomjonovna Ruzieva, Umedjon Amrillojonovich Rustamov Bukhara University of Innovation, rector@bui.uz, umid.rustamov@bui.uz Annotation: Most programming environments used in education today do contain built-in code review features, but feedback quality might not be what it could be. In the current COVID-19 settings, face-to-face code review, or pair programming exercises, is almost entirely gone. A multi-faceted approach to evaluating code, perhaps combining human and AI generated review, rather than an explicit call for an implementable solution would be more appropriate. Future applications could also combine user generated and LLM generated feedback, showing human reviewers the source of the feedback during their reviews [10]. With that goal in mind, this study mined (1) the existing ways that teachers evaluate code quality in order to derive rubric items; (2) the existing ways that rubric items are operationalized into scoring strategies; and (3) the data associated with code submitted as part of a common CS0 curricular item for the purposes of providing a set of student solutions and a giant word tree. Keywords: Source Code Plagiarism (SCP) Detection; Intelligent Coding Education System; Code Review Automation; AI-Driven Personalized Feedback; Large Language Models (LLMs); Token-Based Analysis; Machine Learning Algorithms; Programming Education; Introduction Learning programming is often experienced as hard by students. It involves mastering the formal programming language syntax, problem-solving skills, computational thinking, and testing/debugging abilities. As a discipline, practical exercises complement the problem-oriented approach in learning programming. Writing their own programs is important practice for students trying to articulate the solution to a given problem. It requires them to implement the algorithm, understand and use the underlying data structures, be acquainted with the most common programming constructs etc [1]. Learning programming requires practice, commonly happening in the form of weekly homework assignments in traditional programming courses. In these settings, automated programming assignment test suites amass a major stake in checking the validity of the produced results. Testing and grading manually a vast amount of students' assignments is a time-consuming task which propagates into long delays until the students actually receive feedback. However, timely feedback is a key driver of learning. Recent work about the AI-enhanced auto-correction of programming exercises has sparked interest in the ability to use AI to shorten the feedback loop considerably how effective GPT-3.5 is at automating the correction task of student programming assignments is inspected. Introductory programming courses at universities often have large class sizes. Manual teacher corrections, usually done via the checking of text documents and running 228

Sun’iy Intellekt va Raqamli Iqtisodiyot SIRI-2025 the code, are not possible to provide in a timely manner to a vast number of students. Similarly, crowdsourcing feedback via peer review is problematic to scale up [2]. For the students reviewing the assignments, objectively validating submission criteria concerns is hard, whilst subjecting the student code to automatic assessment typically fails to provide a personalized review of the quality of the solution’s logic and implementation, instead concentrating mainly on passing cases generated by a robust solution trying to mirror the task’s description. Overview of AI in Education Involvement of AI in the end-to-end educational process is investigated; the specific research problems in both the planning and knowledge delivery phase are highlighted; the technological progress in addressing those problems in the past two decades is reviewed. Both proactive and reactive engagement of Artificial Intelligence methodologies in an educational system are investigated so as to automate the information packaging, knowledge delivery, and feedback collection for the educational process. In education, there are three different interconnected scopes, namely curriculum development, knowledge delivery, and student evaluation or learner examination, which correspond to the planning, teaching, and assessment phases, respectively. The essential educational objectives, the roles of instructors or teachers as well as the audience, rationales about the teaching content and the educational format, expectations for the examination and learning outcomes, and future improvements of education can be realized through a comprehensive understanding of the addressed scopes: curriculum development (CD), knowledge delivery (KD), and student evaluation (SE or examination or assessment, hereafter named examination) [3]. The investigation of how AI can be proactively and reactively involved in education is very few and critical; the utility of automatic or AI-based consideration is examined so as to aid the presentation of educational content, the prediction of student performances, and the understanding of audience concerns. A taxonomy and research work are done about the involvement of AI in the CD, KD, and SE phases of education. By diagnosing the involvement of AI in education and by investigating the recent progress, the technical clerkship and interested candidates can be educated about the potential development of intelligent educational systems. Consequently, AI can be involved to reduce the excessive burden on human teachers, increase the interest of K-12 pupils, and automate advantages so as to promote creative thinking and problem-solving skills, to benefit the educational process. Personalization in Learning Systems Personalized teaching in learning systems provides entirely adaptive teaching according to the individual differences of students. For a long time, teaching content or video channels that broadcast this content for all students have been popular. Also, they have been unable to advance on this system that operates in one size for everyone. Large education entities are only available to well-off people. Such entities cannot be easily approached by the majority of the population. Because of that, while enriched educational contents were earlier presented before some parts of the society, their portions reached to more people as of the start of broadcasting. But with the technology, while even more extended educational contents are easier to reach, these cannot be consumed greedily. The 229

SIRI-2025 Sun’iy Intellekt va Raqamli Iqtisodiyot personalized teaching method takes the first positivist steps to implement artificial intelligence that will make this teaching form workable and applicable. Due to the poor students can have a good education, therefore they have high performance. There might be a variety of educational backgrounds and learning abilities in the same group of learners. They can learn methods at different times and with different success levels. This method is direct teaching to students and develops in the paradigm of non-customized teaching. Given recent advances in information technology and artificial intelligence, webbased education systems have become viable alternatives to traditional classroom teaching. The popularity of these online education systems stemmed from their ability to make high quality education available to a large, and often distant, demographic, at a substantially lower cost than the traditional classroom setting. However, these systems do not take advantage of the personalization that is inherently made possible with web-based education, and continue to be one-size-fits-all. Personalizing the education is much more challenging than this one-size-fits-all paradigm: students need to be provided with personalized teaching according to their knowledge and ability, one must learn the best teaching method for each specific student and task, and finally, the teaching should be adapted online based on feedback of the students. A personalized online education system is described that takes into account the above challenges and provides an automated methodology for designing a personalized education system for a fixed set of teaching methods. By using a novel, nonlinear, multivariate process to describe the interactions between the students and the system, the methodology simultaneously optimizes the sequence of teaching actions and the teaching strategy. AI Technologies for Code Review AI-driven personalized code review systems are an emerging educational technology that leverages artificial intelligence (AI) and machine learning (ML) technologies to allow students to improve their programming skills through interactive, fine-granular, feedback regarding their code quality. Such systems can supplement traditional classroom code reviews, in which instructors provide groups of students with high-level feedback on code projects prior to grading. AI technologies, mainly comprising of AI chatbots, computer vision, and data-driven code review techniques, have been equipped in modern code review tools to collaborate on code development and improve review efficiency. According to this development, a 3DKPO framework is envisioned to systematically organize a set of AI-driven personalized code review systems to improve student learning of K-12 coding courses in the future. Finally, the potential benefits and challenges of such development are further discussed in relation to their adaptivities to software development education [4]. 4.1. Natural Language Processing A basic skill development in software development is writing code. There is wide variety of educational interventions educators may use to help learners gain programming skills, among them the Correction of mistakes in code by formative assessments. One such practice is to manually inspect code written by learners and offer feedback. As an alternative, teachers may ask learners to review each other’s code, this peer review can also be done in the classroom or through assignment submission and feedback can be collected at scale. This traditional review-based method lacks breadth due to limitations on the 230

Sun’iy Intellekt va Raqamli Iqtisodiyot SIRI-2025 number of review items. To address these gaps in the literature, AI-driven personalized code review systems are proposed such that teachers can make more informed decisions about how to allocate their feedback time. Educational research and practice increasingly relies on text-based responses. However, the sheer volume of text data generated by educational systems leads to difficulties in providing formative feedback effectively. Research has been exploring the solution by using Natural Language Processing to facilitate automatic filtering, assessment, or just-in-time feedback of learner-generated content, such as open-ended responses and reflective writings. Yet, to the best of the knowledge, evidence synthesis has not been conducted to explore what computational approaches have been widely investigated for feedback provision among post-secondary learners, and the research design, underlying learning theory, and evaluation methods reported in the literature, e.g., large-scale validation or case study involving longitudinal data are insufficient [5]. The extant community cannot make well-informed decisions about the intervention of various types of computational approaches better than others given the current state of evidence. Moreover, it is unclear which important considerations need attention when designing and implementing an intervention across the field. 4.2. Machine Learning Algorithms With the increasing number of digital tools used to create and assess coding exercises, educators have new tools to support students during the programming process and to assess students’ programming exercises. Moreover, computer systems can keep track of student activities, such as code produced, time necessary to solve a task, and the number of attempts to solve it, achieving a better understanding of the process, of the type of mistakes made and the effort required. Recently, AI technologies have reached a remarkable maturity level, providing programmers with a collection of sophisticated models and ready-to-use applications. For these reasons, there has been an increasing research effort aimed at studying the interaction between AI tools and software development education [1]. Researchers have created a new personal AI (PAI) model to boost support for students studying introductory computer science concepts. The AI code review involves the creation of an AI code snippet analysis technique related to a set of code snippets that can be used in college programming education. The AI code comment generation is proposed to review any code snippet provided by the student within the scope of a given subject and provide a review comment for the code snippet based on the analysis of the AI code comment generation enabled technique. A learning platform includes a PRS server for analyzing and providing compiled results after parsing the programs submitted by students, a database storing problem assignments, data of available code snippets, and answer models which corresponds to the answers to the problem assignments [6]. A machine learning model trained on a large labeled set assigns each code snippet to a plurality of predefined categories representing possible comments on the code snippet. The compiled results and the matches between the code snippet and a model answer may be sent to a proprietary cloud server to generate the review comment and display the suggestion to a real-time chat channel. 4.3. Automated Feedback Mechanisms 231

SIRI-2025 Sun’iy Intellekt va Raqamli Iqtisodiyot While the reviews provided by lecturers are invaluable, there exists a substantial resource of time-saver hints and comments that can come only from (and are still very welcome from) peers [7]. Scaffolding the construction of social feedback can be seen as a combinatorial optimization problem, with objectives to maximize utility and learnableness. The texthexnovel randomized search heuristic for generating feedback messages in response to specific programming errors. By having predefined programming goals, students are scaffolding the construction of a solution to create a partial implementation of the process. In software engineering or programming courses, scaffolding is not just limited to the assignment materials. To answer different questions, there are different ways that scaffolding can be employed. One of the more common ways used to guide students towards tasks involves providing several bullet-point hints to nudge them in the right direction. The ability to complete the scaffolding assignment materials can be somewhat expected, and so very briefly guides students with simple hints. The hints provided so far are as simple as possible, modeling the idea that really stuck students will need something more step-by-step. The way that students may gravitate towards requesting help in person rather than reading provided comments is also considered. This is a heuristic setting a time limit for the last hint before the feedback response approach. There are quite a few ways of getting to the same output shape by the time the code executes. This can also manifest with different formats (for example, an idea can be read into a list but then written back as a dictionary). Here, for most questions with multiple desired outputs the student just needs to write a sorted one, but sometimes that needs to be explicitly explained. Instead of restructuring, optionally put all "synonyms" of a task into a correctly formatted dictionary or list [8]. Just as it helps to anticipate comments where they’ve overlooked something obvious, and make sure they’re covered by one of the questions that’s asked. Impact on Learning Outcomes Learning programming is a transformative skill that enhances problem solving, computational thinking, and the ability to design and build software solutions. However, mastering this skill is often experienced as challenging by students, possibly due to the wide range of abilities which need to be developed concurrently. In addition to learning the programming language syntax, students must master problem solving strategies, testing and debugging abilities, coding conventions, and best practices. Learning programming is a major focus of Computer Science curriculums worldwide, which typically require students to practice writing code in the context of labs and homework assignments. Thus, timely feedback on the quality of code produced and overarching design decisions becomes an important driver of learning [2]. In the majority of cases, lecture materials are typically not sufficient for students to grasp the nuances of the assignment at hand. More advanced programming topics often require one-on-one assistance and detailed explanation, which is difficult to provide in crowded classrooms. Students often struggle while mastering what are perceived to be basic programming concepts, and the inability of human teachers to allocate adequate time to address this leads to learners becoming discouraged. To alleviate some of this burden, educators frequently turn to e-assessment systems to expedite the process of grading and feedback giving. A myriad of such systems are 232

Sun’iy Intellekt va Raqamli Iqtisodiyot SIRI-2025 available, providing educators with a wide array of tools to design homework assignments, and auto-grade student submissions. However, the quality of the feedback generated by these systems is typically quite limited, being largely focused on functional correctness in a black-box manner. While recent efforts have made significant strides towards improving the quality of such systems, the majority focus on delivering feedback at the granularity of functions or methods [1]. However, for early students who have yet to grasp more complex programming structures, it is typically necessary to give feedback on statements or code blocks, a task that current e-grading systems are ill-equipped to handle. Meanwhile, design quality is relegated to simple style-checkers. 5.1. Student Engagement Learning to program is often experienced as hard by students, with many dropping out of introductory programming courses. However, programming skills are associated with employability and needed in many STEM disciplines. Timely feedback has been identified as a key driver of learning, yet providing it to students might be increasingly challenging with the growing numbers of students seeking coding education. Crowdsourcing feedback through peer review has been shown to be successful in such settings but might still be limited compared to automated systems. Such information will be helpful for various stakeholders not only in software development companies but in education since autonomous systems taking over quality assurance tasks and evaluating work performance increasingly penetrate many knowledge domains. Major e-assessment systems have been developed that (semi-)automatically assess students’ submissions and return feedback. However, functionality that these systems or learning platforms offer with regard to programming tasks usually focuses solely on functional correctness. There are various smart and automated programming assignment generators that aim to test different programming skills and offer immediate feedback to students. Feedback is either generated automatically or it is manually curated and stacked with problem solutions. In such a way, students can practice programming on a multitude of problems with various levels of difficulty, syntax constraints, etc. There are several studies that analyze how programming assignments can be designed to improve the feedback relevance in a constructive way and to nudge student learning. More recently, powerful AI technologies have emerged which are becoming a game changer in many aspects of society, thus opening up a new field of opportunities in education and learning research. Since the release of certain AI models, many studies have explored possibilities in the domain of educational technology as well, hence focusing not only on text completion but on understanding and generation of programming code. With such systems available to a greater public, a study is conducted that explores the advantages of an AI-assisted approach at scale in the domain of introductory programming education. In the proposed setting, the instructions for a programming assignment and the student solutions are sent to the AI model to mine the tailored feedback. Through different analysis, capabilities and awareness of an AI-based model to assess student-programming task instructions and student-solution pairs for programming instruction are presented. 5.2. Skill Development 233

SIRI-2025 Sun’iy Intellekt va Raqamli Iqtisodiyot Mastering programming by students is often very hard to achieve. This cognitive task involves several difficulties. Among those challenges, students need to understand and use a formal programming language's syntax. They are also required to develop problem-solving skills, implement algorithmic / computational thinking concepts, modularize their code, and fully test and debug their implementation. Testing and debugging are acquirer and refining skills; young coders sometimes feel confident that what is implemented is correct. As a result, they perform few, inaccurate tests — if any — which prevents them from adequately identifying code’s errors. To try to alleviate these issues, the use of an automated eassessment system to grade programming exercises is proposed. Automating coding assessments offer a twofold advantage. On the one hand, students receive immediate and tailored feedback. On the other hand, the time saved by instructors not having to manually grade complex assignments can be invested in creating higher quality pedagogical materials [1]. 5.3. Feedback Quality In the acquisition of software-development skills, feedback that pinpoints errors and explains means of improvement is important in achieving a good student learning experience. However, in both educator and student experiences tutoring and taking programming courses, consistent, helpful feedback is not the norm. It is not feasible to manually provide timely, consistent, and helpful feedback for large or complex coursework tasks. In software coursework, feedback run the risk of being very delayed, simply consisting of grade codes or marks and generally brief or vague qualitative comments [7]. While tools exist to provide feedback to student submissions, their automation is typically limited to reporting either test pass or failure or generating feedback to very simple programming tasks. Tutors, therefore, must still be heavily involved in marking all but straightforward questions and either write out explanations of every bug or every correct code requirement that is not met by the student submission or sit down with students to go through feedback in person, which is unrealistic given the number of submissions. However, anecdotal experience indicates that clusters of students tend to make similar mistakes and/or successes within their coursework. This results in tutors writing out the same feedback, at the same point, for many assignments. Counsell found that students had a tendency to write the same comments if certain mistakes were present in their partner’s code. To help answer these questions, Howell and Wei recently examined feedback given to coursework submissions to a UK level 5, university-level, Data Structures and Algorithms course to determine heuristics used to trigger particular feedback comments that are common between submissions and cohorts, and results are reported. How these heuristics may be used to promote the timeliness and consistency of feedback without jeopardising the quality is also discussed. Conclusion In this article it is argued that now is the perfect time for software engineering educators to embrace AI as a scalable solution to facilitate student's learning in education in a programming course. The graduations of large language models (LLMs) and the availability of code proficient versions of them with Python interfaces have prompted the development of the Utmost AI Observer code review tool [9]. The operations of this more 234

Sun’iy Intellekt va Raqamli Iqtisodiyot SIRI-2025 comprehensive code review tool and the results of two demonstrated studies showing their effectiveness at attracting software development student’s businesses, and field, to a university’s coding contest are described. For the first demonstrated study, university students were invited to code competitively as a part of the on-campus coding contest. It was found that the invitation that their code would be reviewed by Observer, a state-of-the-art AI, written in 73% of invitations, increased the participants by more than 50% compared to inviting them to code competitively with no incentives or implying that their code would be reviewed by experts [1]. Both experimental groups and their firms were ranked substantially higher in the contest participants and on average this same invited-ai group contains more and more participants of the 1st tier compared to the participants of the control group. These same invited-ai finalists have won more and improved more contests and (range) signals from ICPC-style competitions since the invitation than control finalists and participants, respectively. For the second demonstrated study, years of data were analyzed to see trends in the number of annual fields registered from previous programming course contests. It was found that doing at least one thing in 2019 dramatically increased the number of fields coming in subsequent years. Action was associated with implementation. Postmortem field analysis also suggests implementation. References 1. Azaiz, I., Deckarm, O., & Strickroth, S. (2023). AI-enhanced Auto-correction of Programming Exercises: How Effective is GPT-3.5?. 2. Walker Orr, J. (2022). Automatic Assessment of the Design Quality of Student Python and Java Programs. 3. Mallik, S. & Gangopadhyay, A. (2023). Proactive and Reactive Engagement of Artificial Intelligence Methods for Education: A Review. 4. Siow, J. K., Gao, C., Fan, L., Chen, S., & Liu, Y. (2019). CORE: Automating Review Recommendation for Code Changes. 5. Gao, R., E. Merzdorf, H., Anwar, S., Cynthia Hipwell, M., & Srinivasa, A. (2023). Automatic assessment of text-based responses in post-secondary education: A systematic review. 6. Tufano, R., Masiero, S., Mastropaolo, A., Pascarella, L., Poshyvanyk, D., & Bavota, G. (2022). Using Pre-Trained Models to Boost Code Review Automation. 7. Howell, R. & H Wong, S. (2019). Making the Most of Repetitive Mistakes: An Investigation into Heuristics for Selecting and Applying Feedback to Programming Coursework. 8. Suhailan, S., Samad Shibghatullah, A., & A. Burhanuddin, M. (2014). A Perspective Of Automated Programming Error Feedback Approaches In Problem Solving Exercises. 9. Nghiem, K., Minh Nguyen, A., & D. Q. Bui, N. (2024). Envisioning the NextGeneration AI Coding Assistants: Insights & Proposals. 10. Van Petegem, C., Demeyere, K., Maertens, R., Strijbol, N., De Wever, B., Mesuere, B., & Dawyndt, P. (2024). Mining patterns in syntax trees to automate code reviews of student solutions for programming exercises. 235

SIRI-2025 Sun’iy Intellekt va Raqamli Iqtisodiyot SIGNALLARNI INTELEKTUAL TAHLIL QILISHNING USUL VA VOSITALARINI TADQIQ ETISH Sh.T.Muhammadiyeva, K.T.Abdurashidova Kommunikatsiya va raqamli texnologiyalar kafedrasi, “University of Management and Future Technologies”, Toshkent, O‘zbekiston. Annotatsiya: Intelektual signalni tahlil qilish — bu signalni tahlil qilish, filtrlash va undan foydali ma'lumotlarni chiqarish uchun zamonaviy matematik modellari va texnologiyalarni qo‘llashni o‘z ichiga olgan ko‘p sohalarni qamrab olgan fan sohasidir. Ushbu jarayon mashina o‘qitish, neyron tarmoqlari, statistik tahlil va tizimlarni identifikatsiyalash kabi bir qancha metodlarni o‘z ichiga oladi, bu esa ma'lumotlardagi prognoz qilish va klassifikatsiya qilish imkonini beradi. Sun'iy intellektning signalni tahlil qilishdagi roli ushbu sohaning samaradorligini sezilarli darajada oshirib, turli sohalarda, jumladan, telekommunikatsiya, tibbiyot va sanoat monitoringi kabi sohalarda yangiliklar yaratishga yordam beradi. Ushbu maqolada signalni intelektual tahlil qilishda qo‘llaniladigan metodlar va vositalar ko‘rib chiqiladi hamda ularning amaliyotdagi ahamiyati va qo‘llanilishi yoritiladi. Kalit so‘zlar: EEG signallari, sun’iy intellect,fourier transformatsiyasi,mashinani o‘rganish,biotibbiy signallar, neyron tarmoqlar. Kirish. Biotibbiyot texnologiyasidagi rivojlanishlar va miyani tushunishimizning ortib borayotgani tufayli [4], 20-asrning o‘rtalaridan boshlab miya murakkabligi o‘rganila boshlandi va so‘nggi yillarda miya haqidagi fan tadqiqotlar uchun dolzarb mavzulardan biri bo‘lib qolmoqda [5]. Miyadagi elektr faollikni tahlil qilish miya fanining asosiy qiziqish sohalaridan biri hisoblanadi [2]. Natijada, elektroensefalogramma (EEG) miya fanini tahlil qilishda muhim vosita bo‘lib, ko‘pincha miya bilan bog‘liq turli tadqiqot sohalarida qo‘llaniladi [3,7]. EEG signallarini aniqlash va tahlil qilish uchun ularning murakkab va nazariy xususiyatlarini chuqur tushunish, shuningdek, berilgan vazifa uchun tegishli xususiyatlarni ajratib olish zarur. Biroq, EEG signallari o‘ziga xos xususiyatlari tufayli jiddiy muammolarni keltirib chiqaradi. Ushbu muammolardan biri ularning shovqin aralashuviga sezuvchanligi bo‘lib, bu past signal-shovqin nisbatiga olib keladi [6]. Bundan tashqari, ularning noxatolik va normal taqsimotga mos kelmasligi ularni an’anaviy signallardan farqlaydi. Shuningdek, yosh, psixologik holat va sinov muhiti kabi individual omillar EEG signallarida sezilarli o‘zgarishlarni keltirib chiqarishi mumkin [6]. Thakor, N.V ta’kidlanishicha, EEG signallarining o‘ziga xos xususiyatlari ulardan aniq vazifalar haqidagi muhim ma’lumotlarni to‘g‘ridanto‘g‘ri ajratib olishni qiyinlashtiradi [9].Shuning uchun, EEG signallarini yaxshiroq tushunish uchun turli signal tahlili metodologiyalarini ishlab chiqish va mashina o‘qitish usullarini o‘rganish zarur [1]. Muayyan vazifalarga oid tegishli ma’lumotlarni EEG signallaridan aniq ajratib olish ularning o‘ziga xos xususiyatlarini diqqat bilan hisobga olishni va murakkab signal tahlili 236

Sun’iy Intellekt va Raqamli Iqtisodiyot SIRI-2025 metodologiyalarini rivojlantirishni talab qiladi. EEG signallarini aniqlash va tahlil qilish miya faoliyatini tushunishni rivojlantirish uchun juda muhimdir [8]. EEG tadqiqot sohasi sifatida yuqori qiziqish uyg‘otayotgani 2016-2022-yillar oralig‘ida Google Scholar, PubMed va Web of Science ma’lumot qidiruv natijalarida aniq ko‘ringan buni 1-rasmda ko‘rish mumkin. Google scholar Pub med Web of science 7000 6000 5000 4000 3000 2000 1000 0 2016 2017 2018 2019 2020 2021 2022 1-rasm. EEG sohasidagi maqolalar Google Scholar, PubMed va Web of Science ma’lumot qidiruv natijalarida Ushbu maqola EEG signallarining umumiy ishlash prinsipi texnikalari haqida matematik formulalar va psevdokodlar bilan batafsil tavsifni taqdim etgan holda yangi hissani qo‘shadi. Bundan tashqari, EEG sohasidagi so‘nggi yutuqlar haqida hisobot beriladi, mavjud muammolar yoritiladi va kelajakdagi tendensiyalar muhokama qilinadi. Ya’ni 2rasmda EEG signallarining kelajakdagi yangicha yondashuvdagi ishlash prinsiplari. 2-rasm. EEG sohasidagi so‘nggi yutuqlari. Xulosa. Ushbu tadqiqot EEG signallarini tahlil qilish usullari va vositalarini o‘rganishga bag‘ishlandi. Tadqiqot natijalari shuni ko‘rsatadiki, EEG signallarini tahlil qilishda zamonaviy usullar va texnologiyalar keng qo‘llanilmoqda. Asosan, EEG signallarining murakkabligi va 237

SIRI-2025 Sun’iy Intellekt va Raqamli Iqtisodiyot shovqinli tabiati tufayli ular bilan ishlashda yuqori aniqlikni ta’minlash uchun ilg‘or algoritmlar va metodologiyalar zarur. Tadqiqotlar shuni ko‘rsatdiki, signal-shovqin nisbatini yaxshilash, individual fiziologik va psixologik omillarni inobatga olish, hamda tahlil usullarini real vaqt rejimida optimallashtirish EEG signallarining sifatini sezilarli darajada oshiradi. Nazariy adabiyotlar va ilgari o‘tkazilgan tadqiqotlar natijalari EEG signallarini tahlil qilishda shovqinni kamaytirish va aniq xususiyatlarni ajratib olish masalalari asosiy muammolar sifatida ko‘rsatiladi. Shu bilan birga, ilg‘or signal tahlil metodlari va mashina o‘qitish algoritmlarining qo‘llanilishi, EEG signallarining tahlilini yanada samarali qilish imkonini beradi. Xususan, neyron tarmoqlari va chuqur o‘rganish texnologiyalari EEG signallaridan to‘g‘ri va samarali ma’lumot ajratib olishda muhim rol o‘ynaydi. Kelajakda EEG signallarini tahlil qilishda neyrotexnologiyalar, sun’iy intellekt va boshqa ilg‘or usullar yordamida yangi imkoniyatlar ochilishi kutilmoqda. Ushbu metodologiyalarni amaliyotda qo‘llash uchun ularning samaradorligini mustahkamlash va tajriba orqali tasdiqlash zarur bo‘ladi. Shuningdek, EEG signallarining real vaqt rejimida qayta ishlanishi va individual holatlarni hisobga olgan holda maxsus tizimlar ishlab chiqish bu sohadagi tadqiqotlarning istiqbolli yo‘nalishlarini belgilaydi. Adabiyotlar 1. Aggarwal, S.; Chugh, N. Review of machine learning techniques for EEG based brain computer interface. Arch. Comput. Methods Eng. 2022, 29, 3001–3020. 2. Da Silva, F.L. EEG and MEG: Relevance to neuroscience. Neuron 2013, 80, 1112– 1128. 3. Da Silva, F.L. EEG: Origin and measurement. In EEG-fMRI: Physiological Basis, Technique, and Applications; Springer: Berlin/Heidelberg, Germany, 2023; pp. 23–48. 4. Fan, J.; Fang, L.; Wu, J.; Guo, Y.; Dai, Q. From brain science to artificial intelligence. Engineering 2020, 6, 248–252. 5. Lane, R.D.; Ryan, L.; Nadel, L.; Greenberg, L. Memory reconsolidation, emotional arousal, and the process of change in psychotherapy: New insights from brain science. Behav. Brain Sci. 2015, 38, e1. 6. Lotte, F.; Bougrain, L.; Cichocki, A.; Clerc, M.; Congedo, M.; Rakotomamonjy, A.; Yger, F. A review of classification algorithms for EEG-based brain–computer interfaces: A 10year update. J. Neural Eng. 2018, 15, 031005 7. Saibene, A.; Caglioni, M.; Corchs, S.; Gasparini, F. EEG-Based BCIs on Motor Imagery Paradigm Using Wearable Technologies: A Systematic Review. Sensors 2023, 23, 2798. 8. Subha, D.P.; Joseph, P.K.; Acharya U, R.; Lim, C.M. EEG signal analysis: A survey. J. Med. Syst. 2010, 34, 195–212 9. Thakor, N.V.; Tong, S. Advances in quantitative electroencephalogram analysis methods. Annu. Rev. Biomed. Eng. 2004, 6, 453–495 238

Sun’iy Intellekt va Raqamli Iqtisodiyot SIRI-2025 АЛГОРИТМЫ ИСКУССТВЕННОГО ИНТЕЛЛЕКТА В АДАПТИВНОМ ОБУЧЕНИИ Бахриева Х.А.1, Мавлянова Н.З.2 1Vatan University university, Ташкент, xurshidabahriyeva@gmail.com 2UMFT магистрантка 1-курса, Tашкент, nodiramavlonova1992@gmail.com Аннотация. В статье рассмотрены возможности адаптивного обучения, которая имеет способности персонализации образовательного процесса, учитывая индивидуальные потребности и особенности каждого обучающегося. Алгоритмы искусственного интеллекта, применяемые в системах адаптивного обучения. Анализирован методы, машинного обучения, искусственные нейронные сети, обработка естественного языка. Особое внимание уделяется на эффективность, точность и применимости в различных образовательных контекстах. Проведен обзор современных исследований, посвященных использованию искусственного интеллекта в образовании, обсуждаются преимущества и ограничения алгоритмов. Проанализировано перспективы использования искусственного интеллекта в адаптивном обучении, которая позволяет значительно повысить качество образовательного процесса, однако требует дальнейшего развития и уточнения для минимизации ошибок, приводит к улучшению адаптации под нужды пользователей. Приведена рекомендации по выбору и применению алгоритмов искусственного интеллекта в образовательных системах, учитывая их производительность и способность адаптироваться к индивидуальным потребностям учащихся. Ключевые слова: адаптивное обучение, искусственный интеллект, машинное обучение, нейронные сети, персонализация, образовательные технологии, анализ данных, эффективность, точность, образовательный процесс, исследования, алгоритмы. Современное образование сталкивается с вызовами, связанными с разнообразием потребностей учащихся, различиями в их способностях и темпах обучения. Традиционные методы обучения зачастую не могут обеспечить персонализированный подход, что приводит к снижению эффективности образовательного процесса. В этот ряд можно отнести адаптивное обучение, которая основанное на использовании технологий искусственного интеллекта, становится ключевым инструментом для создания индивидуальных образовательных траекторий. С появлением таких технологий как искусственный интеллект, нейронные сети, машинное обучение и анализ больших данных, открывается ряд возможностей по цифровой трансформации сферы образования. При этом применения алгоритмов искусственного интеллекта кардинально меняет парадигму и концептуальную основу образовательной системы, подняв качество образования на новый уровень [1]. 239

SIRI-2025 Sun’iy Intellekt va Raqamli Iqtisodiyot На сегодняшней день на первый уровень выходит персонализация образовательного процесса с учетом потребностей и возможностей обучающегося человека. Существует обучению: несколько основных подходов к персонализированному − Адаптивное обучение: технология, используемая для распределения человеческих или цифровых ресурсов между учащимися в зависимости от их уникальных потребностей. − Индивидуализированное обучение: темп обучения корректируется в соответствии с потребностями отдельных учащихся. − Дифференцированное обучение: подход к обучению корректируется в соответствии с потребностями отдельных учащихся. − Обучение на основе компетенций: учащиеся продвигаются по пути обучения на основе их способности продемонстрировать компетентность, включая применение и создание знаний, а также навыков и предрасположенностей. Последние годы исследования в области применения элементов искусственного интеллекта в образовании показывают свою эффективность и дает положительный результат. Адаптивное обучение — это подход, который использует данные о поведении и успеваемости учащихся для динамической адаптации учебных материалов и методов обучения. Искусственный интеллект играет важную роль в этом процессе, позволяя анализировать большие объемы данных, выявлять закономерности и прогнозировать потребности учащихся. Алгоритмы искусственного интеллекта, такие как машинное обучение, искусственные нейронные сети, метод обратного распространения и рекомендательные системы, генетические, квантовые алгоритмы активно применяются в образовательных платформах для повышения эффективности обучения. Цель данной статьи — провести анализ возможностей основных алгоритмов искусственного интеллекта, используемых в адаптивном обучении, и оценить качества эффективности использования данного подхода. В статье рассматриваются преимущества и ограничения каждого метода, кроме этого, обсуждаются перспективы их применения в образовании. Особое внимание уделяется вопросам точности, адаптивности и персонализации, которые являются ключевыми для успешного внедрения искусственного интеллекта в образовательный процесс. Персонализированное обучение – На основе этой технологии лежит индивидуализация учебного процесса и направленность образовательных ресурсов на одного человека исходя из компетенции, уровня знания студента, а также других когнитивных показателей. Персонализированное обучение «в чистом виде» применяется очень редко. Так как человек — это часть общество, и он должен взаимодействовать с другими 240