Sun’iy Intellekt va Raqamli Iqtisodiyot

1-jadval Tur Kiberfiribgarlikning asosiy turlari va kiber savodxonlik bilan bog‘liqligi Kanal Asosiy maqsad Kiber savodxonlik bilan bog‘liqligi Fishing Elektron pochta, Login va parollarni veb o‘g‘irlash Soxta saytni aniqlash ko‘nikmasi Smishing SMS xabar Bank ma'lumotlarini Noma'lum havolalarga ishonmaslik o‘g‘irlash Vishing Telefon, ovoz Shaxsiy va moliyaviy ma'lumot olish Soxta qo‘ng‘iroqni tanish qobiliyati Onlayn savdo E-commerce firibgarligi platformalar To‘lov va tovar firibgarligi Soxta savdo saytlarini aniqlash Identifikatsiya Turli raqamli o‘g‘irligi kanallar Shaxsni noqonuniy o‘zlashtirish Shaxsiy ma'lumot himoyasi bilimi Ransomware Elektron pochta, Ma'lumotlarni yuklab olish bloklash va tovlamachilik Zararli dasturlarni aniqlash ko‘nikmasi Ijtimoiy tarmoq firibgarligi Ijtimoiy tarmoqlar Ijtimoiy muhandislikni tanish Soxta profil orqali aldash Jadvaldan ko‘rinib turibdiki, kiberfiribgarlikning har bir turi foydalanuvchidan muayyan bilim va ko‘nikmani talab qiladi. Fishing, smishing va vishing bir xil maqsadga - foydalanuvchini aldab maxfiy ma'lumot olishga qaratilgan bo‘lsa-da, ular turli kanallar (elektron pochta, SMS, telefon) orqali amalga oshiriladi va shu sababli ilmiy adabiyotlarda alohida tasniflanadi [6; 7]. Quyida har bir tur batafsil ko‘rib chiqiladi. 1.1. Fishing (Phishing) Fishing - foydalanuvchilarni soxta veb-resurslarga yo‘naltirish va psixologik hamda ijtimoiy muhandislik usullari orqali ulardan shaxsiy, korporativ va autentifikatsiya ma'lumotlarini oshkor ettirishga qaratilgan axborot xavfsizligiga jiddiy tahdid. Jinoyatchilar avval soxta veb-sayt yaratadilar, so‘ngra unga olib boruvchi havolalarni elektron pochta yoki ijtimoiy tarmoqlar orqali tarqatadilar; foydalanuvchilar zararli resurslarga yo‘naltiriladi va maxfiy ma'lumotlarini oshkor qiladi [6]. Kiber savodxonligi yuqori foydalanuvchi esa URL manzilini tekshirish, SSL sertifikatini aniqlash va shubhali so‘rovni tanib olish orqali bunday hujumni oldini ola oladi. 1.2. Smishing Smishing - "SMS" va "phishing" so‘zlarining qo‘shilishidan hosil bo‘lgan atama bo‘lib, mobil qurilma foydalanuvchilariga firibgarona SMS-xabarlar yuborish orqali amalga oshiriladi. Hujum asosan ijtimoiy muhandislik tamoyillariga asoslanib, odamlardagi qo‘rquv, shoshilinchlik va stress kabi psixologik instinktlardan foydalanadi [7]. Tajovuzkor o‘zini bank vakili sifatida taqdim etib, foydalanuvchini soxta havola orqali autentifikatsiya ma'lumotlarini 339

kiritishga undaydi. Kiber savodxonligi past foydalanuvchi SMS orqali kelgan havolaga ishonib, bank kartasi ma'lumotlarini yo‘qotishi mumkin. 1.3. Vishing (Ovozli fishing) Vishing (voice phishing) - kiberjinoyatchilar telefon qo‘ng‘iroqlari yoki ovozli xabarlar orqali foydalanuvchilarni aldab, bank hisob raqamlari yoki kirish ma'lumotlari kabi maxfiy axborotni olishga harakat qiladigan ijtimoiy muhandislik hujumining bir turi [8]. Tajovuzkorlar ko‘pincha bank yoki davlat muassasalari vakillari sifatida o‘zlarini taqdim etadi va qurbonni shoshilinch qaror qabul qilishga majbur qiladi [9]. Bunday hujumni aniqlash uchun foydalanuvchi rasmiy raqamlarni mustaqil tekshirish va telefon orqali maxfiy ma'lumot bermaslik qoidasini bilishi zarur. 1.4. Onlayn savdo firibgarligi Onlayn savdo firibgarligi- soxta elektron tijorat platformalari yoki e'lonlar orqali xaridor yoki sotuvchini aldash orqali pul yoki tovar o‘g‘irlashga qaratilgan kiberjinoyat turi. Jinoyatchilar soxta do‘konlar, noto‘g‘ri tovar tavsifnomalari yoki to‘lovni olganidan keyin tovarni yetkazib bermaslik usullaridan foydalanadi [10]. Kiber savodxonligi yetarli foydalanuvchi savdo platformasining ishonchliligini tekshirish, foydalanuvchi sharhlarini o‘qish va xavfsiz to‘lov usullarini afzal ko‘rish orqali bunday firibgarlikdan himoyalanishi mumkin. 1.5. Identifikatsiya o‘g‘irligi (Identity theft) Identifikatsiya o‘g‘irligi - jinoyatchining boshqa shaxsning shaxsiy ma'lumotlari (pasport raqami, PINFL, bank rekvizitlari)ni noqonuniy ravishda egallab, undan moliyaviy manfaat ko‘rish yoki qonunbuzarlik sodir etish maqsadida foydalanishi. Bu turdagi kiberfiribgarlik odatda fishing, ma'lumotlar bazasi buzilishi yoki ijtimoiy muhandislik orqali olingan ma'lumotlar asosida amalga oshiriladi [10]. Kiber savodxonligi yuqori foydalanuvchi o‘z shaxsiy ma'lumotlarini himoya qilishning muhimligini tushunadi, keraksiz resurslarga ma'lumot qoldirmaydi va kredit tarixini muntazam kuzatadi. 1.6. Ransomware (Tovlamachilik dasturlari) Ransomware - zararli dasturning bir turi bo‘lib, u qurilmadagi fayllarni shifrlaydi va ularni tiklash evaziga to‘lov talab qiladi. Bunday dasturlar odatda fishing elektron pochtalari, zararli ilovalar yoki ishonchsiz manbalardan yuklab olingan fayllar orqali tarqaladi [11]. Kiberjinoyatlarning bu turi nafaqat shaxslarga, balki tashkilotlar va davlat idoralariga ham jiddiy zarar yetkazadi. Kiber savodxonligi yuqori foydalanuvchi noma'lum manbalardagi ilovalarni yuklab olmaslik, dasturlarni muntazam yangilash va ma'lumotlarni zaxiralash ko‘nikmalariga ega bo‘ladi. 1.7. Ijtimoiy tarmoq firibgarligi Ijtimoiy tarmoq firibgarligi - soxta profil yoki buzilgan hisob qaydnomalari orqali foydalanuvchilarni aldash, ulardan pul so‘rash yoki shaxsiy ma'lumot olishga qaratilgan kiberjinoyat turi. Jinoyatchilar ko‘pincha qurbon tanishlarining profilini soxtalashtiradi va shoshilinch moliyaviy yordam so‘rash yoki soxta o‘yin/konkursga jalb qilish usullaridan foydalanadi [10]. Bu turdagi firibgarlikdan himoyalanish uchun foydalanuvchi ijtimoiy tarmoqlardagi so‘rovlarni tanqidiy 340

baholash va shaxsiy ma'lumotlarning oshkoraligini cheklash ko‘nikmalariga ega bo‘lishi zarur. Yuqorida ko‘rib chiqilgan barcha turdagi kiberfiribgarlikni tahlil qilish shuni ko‘rsatadiki, ularning barchasida umumiy bir zaiflik mavjud, jumladan foydalanuvchining raqamli muhitdagi hushyorligi va bilim darajasining yetarli emasligidir. Aynan shu zaiflikni bartaraf etishda kiber savodxonlik hal qiluvchi rol o‘ynaydi. 1.8. Kiberfiribgarlikning oqibatlari va xavf turlari Kiberfiribgarlik faqat texnik muammo bo‘lib qolmay, balki shaxs, jamiyat va davlat uchun jiddiy oqibatlar keltirib chiqaradi. Bu oqibatlarni to‘liq tushunish kiberfiribgarlikka qarshi kurashning strategik muhimligini asoslash uchun zarur. Ilmiy adabiyotlarda kiberfiribgarlikning oqibatlari uch asosiy yo‘nalishda ko‘rib chiqiladi: moliyaviy zarar, psixologik ta'sir hamda reputatsiya va ijtimoiy zarar [10; 12]. Moliyaviy zarar. Kiberfiribgarlikning eng bevosita va o‘lchanadigan oqibati moliyaviy yo‘qotishdir. O‘zbekistonda 2020–2024-yillar oralig‘ida kiberjinoyatlar natijasida yetkazilgan umumiy moddiy zarar 2 trillion so‘mdan oshib ketdi [1]. Jahon miqyosida esa Cybersecurity Ventures tadqiqotiga ko‘ra, kiberjinoyatlardan yetkazilgan zarar 2025-yilga kelib yillik 10,5 trillion AQSh dollariga yetishi bashorat qilingan [12]. Moliyaviy zarar nafaqat bevosita pul yo‘qotish (bank hisobidan mablag‘ o‘g‘irlanishi, ransomware to‘lovi), balki huquqiy xarajatlar, tizimlarni tiklash va ish to‘xtab qolishi oqibatidagi bilvosita yo‘qotishlarni ham o‘z ichiga oladi [10]. Psixologik ta'sir. Tadqiqotlar shuni ko‘rsatadiki, firibgarlikka uchragan shaxslarda tashvish, stress, uyat va o‘ziga bo‘lgan ishonchning yo‘qolishi kabi holatlar keng tarqalgan [13]. Ayniqsa, identifikatsiya o‘g‘irligi va onlayn savdo firibgarligi qurbonlari ko‘pincha uzoq muddatli stress va raqamli xizmatlardan foydalanishdan qo‘rqish holatini boshdan kechiradi. Bu esa pirovardida raqamli iqtisodiyotda faol ishtirokni kamaytiradi va jamiyatning texnologiyalarga bo‘lgan ishonchini susaytiradi [13]. Reputatsiya va ijtimoiy zarar. Kiberfiribgarlik shaxsning ijtimoiy obro‘e'tiboriga ham jiddiy zarar yetkazishi mumkin. Identifikatsiya o‘g‘irligi holatida jinoyatchi qurbonning nomidan noqonuniy faoliyat amalga oshirishi, bu esa shaxsning ish, kredit yoki ijtimoiy munosabatlardagi ishonchini yo‘qotishiga olib kelishi mumkin [10]. Tashkilotlar uchun esa ma'lumotlar buzilishi nafaqat moliyaviy, balki mijozlar ishonchini yo‘qotish va bozor ulushining kamayishi kabi uzoq muddatli reputatsion zararni ham keltirib chiqaradi [12]. Shunday qilib, kiberfiribgarlikning oqibatlari ko‘p qirrali va o‘zaro bog‘liq bo‘lib, ular moliyaviy yo‘qotish psixologik stressga, psixologik stress esa raqamli xizmatlarga bo‘lgan ishonchning pasayishiga olib keladi. Bu o‘zaro bog‘liq ta'sirlar zanjiri kiberfiribgarlikni oddiy jinoyatdan ko‘ra kengroq ijtimoiy muammoga aylantiradi va uni faqat texnik emas, balki ta'limiy va ijtimoiy choralar orqali ham hal etish zarurligini ko‘rsatadi [12; 13]. 2. Kiber savodxonlik: ta'rif va komponentlar 341

Kiber savodxonlik (cyber literacy) - bu shaxsning raqamli muhitda xavfsiz, mas'uliyatli va samarali tarzda harakat qila olish qobiliyatini ta'minlovchi bilimlar, ko‘nikmalar va xatti-harakatlar majmuasi [14]. Bu tushuncha faqat texnologiyadan foydalana olish ko‘nikmalarini emas, balki raqamli tahdidlarni tanib olish, ulardan himoyalanish va tegishli qaror qabul qilish kompetentsiyasini ham qamrab oladi. Ilmiy adabiyotlarda kiber savodxonlik ko‘pincha "raqamli savodxonlik" (digital literacy) bilan bir qatorda ko‘rib chiqilsa-da, ular o‘rtasida muhim farq mavjud. Raqamli savodxonlik texnologiyadan umumiy foydalanish ko‘nikmalarini, kiber savodxonlik esa xavfsizlik va xatarlarni boshqarish aspektlarini alohida ta'kidlaydi [14]. 2.1. Kiber savodxonlikning asosiy komponentlari Tadqiqotchilar kiber savodxonlikni uchta asosiy komponent orqali tavsiflaydi. Vishwanath va boshqalar (2020) [14] kiber savodxonlikni bilish-tushunish (kognitiv), ko‘nikma (xulq-atvor) va motivatsion komponentlarga ajratadi. Canham va boshqalar (2021) [15] esa kontekstga oid bilim (kontekstual savodxonlik) muhimligini alohida ta'kidlaydi - hujum usullari faqat umumiy emas, balki vaziyatga mos holda tanilishi zarur, deb hisoblaydi. – Bilim (Knowledge): Kiberxavflar, fishing, zararli dasturlar, maxfiylik qoidalari va axborot xavfsizligi tamoyillari haqida nazariy ma'lumotga ega bo‘lish [14]. – Ko‘nikmalar (Skills): Shubhali xabar yoki veb-saytni aniqlash, kuchli parol yaratish, ikki bosqichli autentifikatsiyadan foydalanish kabi amaliy qobiliyatlar [15]. – Xatti-harakat (Behavior): Onlayn faoliyatda ehtiyot bo‘lish, noma'lum havolalarni bosmaslik, shaxsiy ma'lumotlarni himoya qilish kabi odatiy xulq-atvor [14]. Ushbu uch komponent o‘zaro bir-birini to‘ldiradi: bilimning amaliyotda qo‘llanilishi ko‘nikmaga, ko‘nikmaning muntazam qo‘llanilishi esa to‘g‘ri xattiharakatga aylanadi [15; 14]. 2.2. Kiber savodxonlik va kiberfiribgarlikka ta'sirchanlik o‘rtasidagi bog‘liqlik Bir qator empirik tadqiqotlar kiber savodxonlik darajasi va kiberxavflarga duchor bo‘lish ehtimoli o‘rtasida teskari mutanosiblik mavjudligini ko‘rsatadi ya'ni, kiber savodxonligi yuqori bo‘lgan foydalanuvchilar fishing hujumlarini ancha samarali aniqlaydi [14]. Vishwanath va boshqalar (2020) [14] tomonidan o‘tkazilgan tadqiqotda fishing hujumlariga nisbatan ta'sirchanliklarda kognitiv omillar - ya'ni axborotni tanqidiy baholash ko‘nikmasi - hal qiluvchi rol o‘ynashi aniqlandi. Canham va boshqalar (2021) [15] tadqiqoti esa shuni ko‘rsatdiki, maqsadli (spear-phishing) hujumlarda shaxsiy ma'lumotlarga asoslangan moslashtirish fishing emailini aniqlashni sezilarli darajada qiyinlashtiradi va bu kontekstga oid savodxonlikning muhimligini tasdiqlaydi. O‘zbekiston misolida esa vaziyat yanada keskinroq bo‘lib, kiberjinoyatlar 2019--2024-yillar oralig‘ida 68 barobarga oshgan [1]. Bu o‘sish nafaqat 342

kiberxavflarning kuchayishi, balki aholining ularga tayyorgarlik darajasi yetarli emasligini ham ko‘rsatadi. Shuning uchun kiber savodxonlikni oshirish bu muammoni hal etishning eng muhim strategik yo‘nalishi hisoblanadi. 3. Kiber savodxonlikni oshirish usullari Kiber savodxonlikni oshirish uchun bitta usul yetarli emas - eng samarali natijaga faqat bir-birini to‘ldiruvchi yondashuvlar kombinatsiyasi orqali erishish mumkin [14]. Shu maqsadda quyidagi uch bosqichli tizim taklif etiladi: 1-rasm. Kiberfiribgarlikni oshirishning uch bosqichli usullari. 3.1. Onlayn kurslar Birinchi bosqich sifatida onlayn kurslar kiber savodxonlikning poydevorini shakllantiradi va kurslar aholining turli qatlamlari - talabalar, davlat sektori xodimlari va xususiy sektor vakillari - ehtiyojlarini hisobga olgan holda ishlab chiqilishi lozim. Onlayn ta'lim platformalari foydalanuvchilarga fishing, ijtimoiy muhandislik, zararli dasturlar va shaxsiy ma'lumotlarni himoya qilish usullari bo‘yicha tizimli nazariy bilim beradi [14]. Ushbu kurslarning asosiy afzalligi shundaki, ular istalgan vaqt va joydan foydalanish imkonini berishi va bilimlarni muntazam yangilab borish imkoniyatini ta'minlaydi. 3.2. Simulyatsiya qilingan fishing hujumlari Nazariy bilimlarning o‘zi yetarli emas - ularni amalda sinab ko‘rish zarur [15]. Shu sababli ikkinchi bosqich sifatida onlayn kurslarni tugatgan foydalanuvchilar uchun simulyatsiya qilingan fishing hujumlari amalga oshiriladi. Foydalanuvchilarga haqiqiy hujumlarni eslatuvchi, ammo hech qanday zarar yetkazmaydigan sinov xabarlari yuboriladi, natijalar tahlil qilinib, maqsadli ta'lim mashg‘ulotlari tashkil etiladi. Tadqiqotlar shuni ko‘rsatadiki, simulyatsiya asosidagi o‘qitish fishing hujumlariga nisbatan ta'sirchanliklini 40--60 foizga kamaytirishi mumkin [14]. 3.3. Ommaviy axborot vositalari orqali ma'rifat 343

Birinchi ikki bosqich faqat kurslarda ishtirok etgan foydalanuvchilarni qamrab oladi, lekin jamiyatning kengroq qatlamlariga yetib borish uchun uchinchi bosqich - axborot vositalari - hal qiluvchi rol o‘ynaydi. Ommaviy axborot vositalari, ijtimoiy tarmoqlar va rasmiy veb-saytlar orqali olib boriladigan ma'rifiy tadbirlar kiberfiribgarlikning eng keng tarqalgan usullari, ularning oqibatlari va ularga qarshi himoya yo‘llari haqida oddiy tilda tushuntirib, jamiyatning "raqamli immunitetini" mustahkamlaydi [1; 2]. Xulosa. Kiberfiribgarlikning oldini olish texnik himoya choralaridan ancha kengroq yondashuvni talab etadi. Ushbu maqolada ko‘rib chiqilgan kiberfiribgarlikning yetti turi - fishing, smishing, vishing, onlayn savdo firibgarligi, identifikatsiya o‘g‘irligi, ransomware va ijtimoiy tarmoq firibgarligi - barchasida umumiy zaiflik sifatida foydalanuvchining bilim va hushyorlik darajasining yetarli emasligi aniqlandi. Bu esa kiber savodxonlikni oshirishni strategik muhim yo‘nalish sifatida belgilaydi [14]. Taklif etilgan uch bosqichli tizim - onlayn kurslar, simulyatsiya qilingan fishing hujumlari va ma'rifiy axborot vositalari - bir-birini to‘ldiradi: kurslar bilim beradi [14], simulyatsiyalar ushbu bilimni mustahkamlaydi [15], axborot vositalari esa xabardorlikni jamiyatning barcha qatlamlariga yoyadi. O‘zbekistonda kiberjinoyatlarning 68 barobarga oshganligi [1] bunday tizimni joriy etishni dolzarb vazifaga aylantiradi. Adabiyotlar ro‘yxati 1. Kun.uz. (2025, May 29). Uzbekistan sees 68-fold surge in cybercrime: Nearly 2 trillion UZS stolen in five years. https://kun.uz/en/news/2025/05/29/uzbekistan-sees-68-fold-surge-incybercrime-nearly-2-trillion-uzs-stolen-in-five-years 2. Kun.uz. (2025, November 3). Uzbekistan steps up fight against drug abuse and cybercrime. https://kun.uz/en/news/2025/11/03/uzbekistan-steps-up-fight-against-drug-abuseand-cybercrime 3. Inlibrary. (2022). O‘zbekiston Respublikasining "Kiberxavfsizlik to‘g‘risida"gi qonuni. https://inlibrary.uz/index.php/science-research/article/view/139462 4. Tashkent Times. (2025, April 30). Uzbekistan to step up fight against cybercrime. https://mail.tashkenttimes.uz/national/15162-uzbekistan-to-step-up-fight-against-cybercrime 5. Zamin.uz. (2023). Why are 98% of cyberattacks targeted at bank cards? https://zamin.uz/en/society/167833-why-are-98-of-cyberattacks-targeted-at-bank-cards.html 6. Alabdan, F., Perumal, S., & Yahia, S. B. (2023). Phishing attacks in the Internet of Things: A systematic literature review. IEEE Access. https://ieeexplore.ieee.org/document/10049452 7. Joo, J. W., Moon, S. Y., Singh, S., & Park, J. H. (2022). S-detector: An enhanced security model for suspicious smishing detection. Telecommunication Systems. https://ieeexplore.ieee.org/document/9931666 8. Group-IB. (2023). Vishing: Voice phishing explained. https://www.groupib.com/resources/knowledge-hub/vishing/ 9. TechTarget. (2023). What is vishing (voice or VoIP phishing)? https://www.techtarget.com/searchunifiedcommunications/definition/vishing 10. Anderson, R. (2020). Security Engineering: A Guide to Building Dependable Distributed Systems (3rd ed.). Wiley. 11. Brewer, R. (2016). Ransomware attacks: Detection, prevention and cure. Network Security, 2016(9), 5--9. https://doi.org/10.1016/S1353-4858(16)30086-1 344

12. Morgan, S. (2020). Cybercrime to cost the world $10.5 trillion annually by 2025. Cybercrime Magazine. https://cybersecurityventures.com/hackerpocalypse-cybercrime-report2016/ 13. Button, M., & Cross, C. (2017). Cyber Frauds, Scams and their Victims. Routledge. 14. Vishwanath, A., Herath, T., Chen, R., Wang, J., & Rao, H. R. (2020). Why do people get phished? Testing individual differences in phishing vulnerability within an integrated, information processing model. Decision Support Systems, 51(3), 576--586. 15. Canham, M., Poeppl, T., Bellekens, X., & Speakman, L. (2021). Spear phishing 101: Understanding the impact of personalization, relevance, and priming on phishing susceptibility. Journal of Cybersecurity & Privacy, 1(2), 350--370. MACHINE LEARNING MODELS FOR BREAST CANCER DETECTION Jaloliddin Omonbayev, Master’s student at Urgench State University named after Abu Rayhan Beruni, jaloliddinomonboev@gmail.com Shokhrukhbek Rakhimboyev, Master’s student at Urgench State University named after Abu Rayhan Beruni, raximboyevshohrux96@gmail.com Annotatsiya. Mazkur maqolada ko‘krak bezi saratonini aniqlashda qo‘llaniladigan mashina o‘rganish modellari chuqur tahlil qilinadi. Modellarning afzalliklari, kamchiliklari va tibbiy tashxis jarayonidagi amaliy qo‘llanilishi ko‘rib chiqiladi. Kalit sozlar: mashina o‘rganish, ko‘krak bezi saratoni, tasniflash, tibbiy tashxis, bashoratlash Аннотация. В статье проводится расширенный анализ моделей машинного обучения, применяемых для выявления рака молочной железы. Оцениваются их достоинства, ограничения и практическая применимость в медицинской диагностике. Ключевые слова: машинное обучение, рак молочной железы, классификация, медицинская диагностика, прогнозирование Abstract. This paper provides an extended analytical review of machine learning models used for breast cancer detection. The advantages, limitations, and practical applicability of classical machine learning algorithms in medical diagnosis are examined to support informed model selection for experimental research. Key words: machine learning, breast cancer, classification, medical diagnosis, prediction Introduction. Breast cancer is a problem for women all over the world. It is one of the reasons why women die from cancer. The World Health Organization and other international health groups say that finding breast cancer early and getting the diagnosis are very important. This helps to reduce the number of women who die from breast cancer and helps them live longer.. The usual ways of diagnosing breast cancer like mammograms, ultrasound imaging and biopsies can take a long time. These methods also depend a lot, on how skilled the doctors and medical professionalsre. Breast cancer diagnosis is an issue and breast cancer patients need a better way to get diagnosed.[1] The last few years have seen a change in the way we keep medical records and diagnostic information. Now that we have access to digital medical records and diagnostic data we can use machine learning in healthcare in new and exciting ways. Machine learning is really good, at looking at amounts of data finding patterns that are not easy to see and helping doctors make better decisions when they are diagnosing patients. When it comes to finding breast cancer machine learning is used to look at tumors and decide if they are benign or malignant. This 345

is done by looking at features of the tumor that doctors and nurses have found. Breast cancer detection is an area where machine learning can be very helpful. Machine learning models are used to classify tumors as benign or malignant based on morphological features of breast cancer. People are getting more interested in diagnostic systems. This shows how important it is to pick the machine learning models. These models need to be accurate, robust and easy to understand. When it comes to applications it is really crucial that the models are transparent. Doctors need to understand and trust the predictions made by these automated systems, like diagnostic systems. They have to be able to rely on diagnostic systems. The objective of this paper is to analyze and compare classical machine learning models used for breast cancer detection, focusing on their theoretical foundations, strengths, and limitations. This analysis aims to provide a solid basis for selecting suitable models in the subsequent experimental phase of the research. The study also contributes to the broader goal of developing reliable computeraided diagnosis systems for clinical practice [1]. Main body. Breast cancer detection widely uses machine learning models. Most of these models rely on supervised learning techniques. So what is supervised learning? It is a type of learning where models are trained using datasets that contain labeled data. Each sample in the dataset has a known outcome or class label. For example, the Wisconsin Breast Cancer Dataset is widely used for this purpose. This dataset contains numerical features that describe the characteristics of a cell nucleus obtained from medical tests. These features help machine learning models learn patterns that distinguish between benign and malignant tumors. As shown in Figure 1, the typical workflow of a breast cancer detection system includes dataset collection, data preprocessing, model training and fine-tuning using a pre-trained model, validation to identify the best-performing model, and finally classification of the results. Breast cancer detection relies on machine learning models, and the Wisconsin Breast Cancer Dataset is an important tool used for this purpose [2]. Figure 1. Supervised learning framework for breast cancer detection Logistic Regression is really useful for looking at data. It helps figure out how likely someone is to have cancer. This is very helpful for doctors when they need to make decisions. The Logistic Regression model also shows which things are most important when looking at the data. But Logistic Regression has a 346

problem. It thinks that things are related in a way. This means it is not good, at understanding patterns when trying to diagnose diseases with Logistic Regression. Support Vector Machines are really good at telling groups apart. They do this by making the space between these groups big as possible. Support Vector Machines can also deal with medical information that is not straightforward. This is because they use tools called kernel functions. Lots of people have used Support Vector Machines to look at breast cancer data. They work really well. Support Vector Machines are very good at diagnosing breast cancer. As shown in Figure 2, different machine learning models create different decision boundaries when classifying data. Logistic Regression produces a linear boundary, while Support Vector Machines can form more flexible boundaries to better separate the classes. Support Vector Machines can be tricky to set up and it is hard to understand how they make their decisions. This is a problem because doctors like to know why a decision was made. So even though Support Vector Machines are good at what they do they might not be used much in hospitals because of these problems, with Support Vector Machines. Figure 2. Decision boundaries of Logistic Regression and Support Vector Machine The k-Nearest Neighbors algorithm is a way of learning that looks at examples that're similar to each other. It is simple and easy to understand, which makes it a good choice when you are first looking at data. The k-Nearest Neighbors algorithm has some problems. It takes a time to make predictions and it does not work well when the data is noisy or not scaled properly. This means the kNearest Neighbors algorithm is not very good for medical projects where you need to get things right every time. The k-Nearest Neighbors algorithm is just not reliable enough, for these kinds of jobs. Decision Tree models are really useful for classifying things. They work by following a set of rules one after the other to make a decision. This makes it easy for doctors and other clinicians to see how the Decision Tree models came up with their predictions, which's very important in healthcare. Decision Tree models have a problem though: they can get too good at fitting the data they are trained on. This is especially true when the dataset is not very big or has some information in it. When this happens the Decision Tree models do not work well on new data so they are not as good, at making predictions. 347

Figure 3. Decision Tree structure and ensemble learning concept Ensemble learning methods, including Random Forest and Gradient Boosting algorithms, combine multiple individual models to improve predictive accuracy and stability. Random Forest mitigates overfitting by averaging the predictions of multiple decision trees, while Gradient Boosting incrementally improves model performance by focusing on misclassified samples. These methods have demonstrated superior results in breast cancer detection tasks, although they introduce increased computational complexity and reduced model interpretability [3]. Table 1 Comparison of machine learning models for breast cancer detection Model Accuracy Interpretability Computational Cost Logistic Regression SVM k-NN Decision Tree Random Forest Gradient Boosting Medium High Low Suitability for Medical Use High High Medium Medium High Very High Low Low High Medium Low High High Medium High Very High Medium Low High High Medium Conclusion. This paper is about breast cancer detection using school machine learning models. It looks at these models closely. What it finds out is that each breast cancer detection model has its good and bad points. This means it is really important to pick the model for the job when it comes to medical stuff. Some methods like methods and Support Vector Machines are really good at figuring out if someone has breast cancer or not. On the hand Logistic Regression and Decision Trees are better, at explaining how they make their decisions and are easier to understand. Breast cancer detection models need to be chosen. In diagnostics it is really important to find a balance between how well a test can predict something and how well we can understand how it works. This is necessary for doctors to start using these tests. The things we learned from this study will help us get ready, for the part of the research. In the part we will try out some models and see how they work with real medical information. We will look at how these models work and compare them to each other. We will also try to find out if we can use computer techniques to help diagnose breast cancer from images [4]. 348

Literature references: 1. Matlatipov, S.G., Rajabov, J., Kuriyozov, E., & Aripov, M. (2024, May). UzABSA: Aspect-Based Sentiment Analysis for the Uzbek Language. In Proceedings of the 3rd Annual Meeting of the Special Interest Group on Under-resourced Languages@ LREC-COLING 2024 (pp. 394-403). 2. Kutlimuratova, B., Kuriyozov, E., Urazbaev, A., & Rakhimova, G. (2025, November). Corpus-Based Error Analysis of Uzbek EFL Learners’ Academic Writing. In 2025 IEEE XVII International Scientific and Technical Conference on Actual Problems of Electronic Instrument Engineering (APEIE) (pp. 1-5). IEEE. 3. Mattiev, J., Davityan, M., & Kavsek, B. (2023). ACMKC: A compact associative classification model using K-modes clustering with rule representations by coverage. Mathematics, 11(18), 3978. 4. Madatov, K., Matlatipov, S., & Aripov, M. (2023). Uzbek text's correspondence with the educational potential of pupils: a case study of the School corpus. arXiv preprint arXiv:2303.00465. TASVIRLI MA’LUMOTLAR ASOSIDA BOSH MIYA SARATONI KASALLIGINI ANIQLASH ALGORITMI VA DASTURIY VOSITASI Qambarov Bahriddin Panji o‘g‘li, TDIU 1-kurs tayanch doktoranti, qambarov.b.p1997@gmail.com Annotatsiya. Ushbu tezisda MRI (magnit-rezonans imaging) tasvirli ma’lumotlar asosida bosh miya saratoni kasalligini aniqlash muammosi ko‘rib chiqilgan. Tadqiqotda YOLOv8s chuqur o‘rganish (Deep Learning) modeli qo‘llanilgan. Model 4 ta turdagi bosh miya saratoni holatini – glioma, meningioma, gipofiz o‘smasi (pituitary), va o‘smasiz holatni real vaqt rejimida aniqlashga qaratilgan. Tajribalar natijalari shuni ko‘rsatadiki, taklif etilgan model mAP50 ko‘rsatkichi bo‘yicha 92.4 % aniqlikka erishdi. Olingan natijalar zamonaviy chuqur o‘rganish modellaridan raqobatbardosh bo‘lib, klinik diagnostikada amaliy qo‘llanilish imkoniyatiga ega. Kalit so‘zlar: miya o‘smasi, YOLOv8, chuqur o‘rganish (Deep Learning), MRT (magnitrezonans tomografiya) tasviri, object detection, bounding box. Аннотация. В данной диссертации рассматривается проблема выявления рака головного мозга на основе данных МРТ (магнитно-резонансной томографии). В исследовании использовалась модель глубокого обучения YOLOv8s (Deep Learning). Модель нацелена на выявление в режиме реального времени 4 типов рака головного мозга — глиомы, менингиомы, опухоли гипофиза и неопухолевого. Результаты экспериментов показывают, что предложенная модель достигла точности 92,4% по показателю mAP50. Полученные результаты конкурентоспособны современным моделям глубокого обучения и имеют возможность практического использования в клинической диагностике. Ключевые слова: опухоль головного мозга, YOLOv8, глубокое обучение, МРТ (магнитно-резонансная томография), обнаружение объектов, ограничивающая рамка. Abstract. This thesis examines the problem of brain cancer detection based on MRI (magnetic resonance imaging) image data. The YOLOv8s deep learning (Deep Learning) model was used in the research. The model is aimed at real-time detection of 4 types of brain cancer glioma, meningioma, pituitary tumor, and non-tumor. The results of the experiments show that the proposed model achieved 92.4% accuracy according to the mAP50 indicator. The obtained results are competitive with modern deep learning models and have the possibility of practical use in clinical diagnostics. Key words: brain tumor, YOLOv8, deep learning, MRI (magnetic resonance imaging), object detection, bounding box. 349

Kirish. Zamonaviy axborot texnologiyalarining jadal su’ratlarda rivojlanishi natijasida obyektlarni tanib olish tizimlari turli sohalarda keng qo‘llanilmoqda. Xususan, bu tizimlar xavfsizlik, tibbiyot, sanoat, transport kabi ko‘plab yo‘nalishlarda muhim ahamiyat kasb etmoqda. Obyektlarni to‘g‘ri va samarali tanib olish algoritmlari orqali inson faoliyatini yengillashtiruvchi, vaqt va resurslarni tejashga xizmat qiluvchi turli amaliy dasturlar yaratilmoqda. Bu jarayon, asosan, kompyuter ko‘zi (Computer Vision) va mashinanili o‘qitish (Machine Learning) algoritmlariga tayanadi. Ayniqsa, Convolutional Neural Network (CNN) algoritmlari bu borada keng qo‘llanilayotgan asosiy texnologiyalardan biridir.[1] Bosh miya saratoni dolzarb tibbiy muammolardan biri hisoblanib, uni erta aniqlash va tashxislash bemorlarning sog‘ayish ko‘rsatkichlarini oshirishda muhim rol o‘ynaydi. Bosh miya o‘smasi inson salomatligi uchun jiddiy xavf tug‘diruvchi, taxminan 200 ga yaqin turli xildagi patologik to‘qima o‘simtalari majmuasidan iborat. Gial hujayralaridan kelib chiquvchi gliomalar xavfli bosh miya o‘smalarining eng ko‘p tarqalgan turi bo‘lib, bunday holatlarning 80% ini tashkil etadi. Jahon sog‘liqni saqlash tashkiloti (JSST) gliomalarni to‘rtta (I-Pilotsitar astrotsitoma, II-Diffuz glioma, III-Anaplastik glioma, IV-Glioblastoma) toifaga tasniflaydi, ularning har biri xavfsiz darajadan o‘ta og‘ir darajagacha bo‘lgan xavflilik spektrini o‘z ichiga oladi.[2] Zamonaviy sun’iy intellekt va chuqur o‘rganish texnologiyalarining rivojlanishi tibbiy rasmlar bilan ishlash va ularni avtomatik tahlil qilish imkonini bermoqda. Ushbu tadqiqotning maqsadi YOLOv8 modeli yordamida MRI tasvirlari asosida bosh miya saratoni kasalligini real vaqt rejimida aniqlash va klassifikatsiyalashdan iborat. Asosiy qism. Ushbu olib borilgan tadqiqot ishda Pkdrabi tomonidan Kaggle platformasida joylashtirilgan “Medical Image Dataset – Brain Tumor Detection” dataseti qo‘llanildi. Dataset YOLO formatida annotatsiyalangan bo‘lib, quyidagi tuzilmaga ega: 1-jadval O‘sma turi Train Validation Test Glioma ~700 ~100 285 Meningioma ~700 ~100 142 Pituitary ~700 ~100 185 Jami ~2100 ~300 612 YOLOv8 (You Only Look Once, versiya 8) Ultralytics kompaniyasi tomonidan 2023-yilda ishlab chiqilgan zamonaviy obyektlarni aniqlash arxitekturasi hisoblanadi. Modelning asosiy afzalliklari: bitta o‘tishda ham obyektni topish, ham klassifikatsiya qilish imkoniyati; yuksak tezlik va aniqlikning muvozanati; transfer learning imkoniyati. YOLOv8 o‘zidan oldingi YOLOv7 modeli bilan solishtirilganda, bosh miya o‘smalarini aniqlashda sezilarli yutuqlarni namoyish etadi; bu asosan yuqori aniqlik, tezkorlik hamda o‘sma turlari va 350

bosqichlarini bir-biridan farqlash imkoniyati bilan bog‘liq. Modelning arxitekturasi real vaqt rejimida ishlash imkonini beradi, bu esa uni o‘z vaqtida tashxis qo‘yish hal qiluvchi ahamiyatga ega bo‘lgan klinik amaliyotlarda qo‘llash uchun qulay qiladi.[3] 1-rasm Ushbu tadqiqotda YOLOv8s (small) modeli tanlangan, chunki u tezlik va aniqlik o‘rtasida optimal muvozanatni ta’minlaydi. Model parametrlari: 11.1 million parametr, 28.4 GFLOPs hisoblash hajmi. Model Kaggle platformasida Tesla T4 GPU yordamida o‘qitildi. O‘qitish parametrlari quyidagicha belgilangan: epochs=50, imgsz=640, batch=16, optimizer=Adam (lr=0.001429), patience=10. O‘qitish jarayoni taxminan 34 daqiqa davom etdi. Xulosa va amaliy tavsiyalar. 50 ta epoch bo‘yicha o‘qitish yakunida YOLOv8s modeli quyidagi natijalarni ko‘rsatdi: 2-jadval O‘sma turi Precision Recall mAP50 mAP50-95 Glioma 0.830 0.719 0.818 0.548 Meningioma 0.957 0.937 0.979 0.829 Pituitary 0.930 0.938 0.975 0.746 Umumiy 0.906 0.865 0.924 0.708 Taklif etilgan YOLOv8s modeli natijalarini adabiyotlarda keltirilgan modellar bilan taqqoslaganda, model raqobatbardosh natijalar ko‘rsatdi:[4, 5, 6] 3-jadval Model Aniqlik (%) F1-score Tezlik EfficientNetV2+ViT 96.0 0.96 — Fuzzy Logic+CNN 98.0 — — DC-YOLOv8FEN 99.5 0.965 — Taklif etilgan (YOLOv8s) 92.4 0.884 16.2ms Jadvaldan ko‘rinib turibdiki, taklif etilgan YOLOv8s modeli 92.4% mAP50 ko‘rsatkichiga erishdi. Glioma turi uchun aniqlik nisbatan past (0.818) bo‘lgan, bu esa gliomaning MRI tasvirlarida noaniq chegaralarga ega bo‘lishligidan kelib 351

chiqadi. Meningioma va gipofiz o‘smasi uchun esa aniqlik 0.975-0.979 darajasida bo‘lib, yuqori natija ko‘rsatdi. Ushbu tadqiqotda YOLOv8s modeli MRI tasvirlarida miya o‘smasini real vaqt rejimida aniqlash uchun muvaffaqiyatli qo‘llanildi. Model 92.4% mAP50 aniqligiga erishib, bitta tasvirni 16.2 millisekunda ichida qayta ishlash imkoniyatini namoyish etdi. Bu natijalar klinik diagnostikada real vaqt rejimidagi tizimlar uchun YOLOv8 arxitekturasining samarali ekanligini tasdiqlaydi. 2-rasm 3-rasm 2-3-rasmlarda YOLOv8s modelining MRI tasvir ustidagi test natijasi ko‘rsatilgan. Model Pituitary o‘smasini 75 % ishonch darajasida aniqlab, chegaralovchi to‘rtburchak yordamida belgiladi. Kelajakda modelning aniqligini oshirish maqsadida quyidagi yo‘nalishlar ko‘rib chiqilishi mumkin: yanada katta annotatsiyalangan datasetdan foydalanish, YOLOv8l yoki YOLOv8x kabi kattaroq model versiyalarini sinab ko‘rish, ma’lumotlarni kuchaytirish (data augmentation) usullarini kengaytirish. Foydalanilgan adabiyotlar ro‘yxati: 1. Timsollarni tanib olishda obyektlar muhimligini axborot o‘lchovi mezonlari va ularni hisoblash algoritmlari. (2025). Conferences, 1(4), 107-113. https://xiuresearch.com/index.php/Conferences/article/view/283 2. Anees Tariq, Muhammad Munwar Iqbal, Muhammad Javed Iqbal , Iftikhar Ahmad (2025). “Transforming Brain Tumor Detection Empowering Multi-Class Classification With Vision Transformers and EfficientNetV2”. https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10945363 3. Lincy Annet Abraham , Gopinath Palanisamy , and Veerapu Goutham (2025). “Dilated Convolution and YOLOv8 Feature Extraction Network: An Improved Method for MRI-Based Brain Tumor Detection”. https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10877809 4. Anees Tariq, Muhammad Munwar Iqbal, Muhammad Javed Iqbal , Iftikhar Ahmad (2025). “Transforming Brain Tumor Detection Empowering Multi-Class Classification With Vision Transformers and EfficientNetV2”. https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10945363 5. Nashaat M. Hussain Hassan, and Wadii Boulila (2025). “ Efficient Approach for Brain Tumor Detection and Classification Using Fuzzy Thresholding and Deep Learning Algorithms”. https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10981713 6. Lincy Annet Abraham , Gopinath Palanisamy , and Veerapu Goutham (2025). “Dilated Convolution and YOLOv8 Feature Extraction Network: An Improved Method for MRI-Based Brain Tumor Detection”.https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10877809 352

GAUSS VA IMPULS SHOVQINLARINI BARTARAF ETISHDA NOCHIZIQLI MEDIAN FILTRLASH USULLARINING SAMARADORLIGI Sultanov Djamshid Baxodirovich, t.f.d., dotsent, email: sdjamshid@gmail.com Narzullayev Inomjon Sadulla o‘g‘li, t.f.f.d. (PhD) email: i.narzullayev@tuit.uz Qodirov Raximjon Rasuljon o‘g‘li, tayanch doktorant Email: Rahimjonqodirov1997uzbek@gmail.com Annotatsiya. Ushbu maqolada raqamli tasvirlarda uchraydigan fotoelektron, Gauss va impuls (tuz va murch) shovqinlari tahlil qilinadi. Tasvir degradatsiyasini kamaytirishda nochiziqli median filtrlash usulining samaradorligi o‘rganilib, uning shovqinni bartaraf etishda yuqori natija berishi ko‘rsatiladi. Kalit so‘zlar: raqamli tasvirlarni qayta ishlash, tasvir shovqini, shovqinni bartaraf etish, nochiziqli filtrlash, tuz va murch shovqini. Аннотация. В статье анализируются фотоэлектронный, гауссовский и импульсный (соль и перец) шумы, возникающие в цифровых изображениях. Исследуется эффективность нелинейной медианной фильтрации при снижении деградации изображения и показана её высокая результативность при подавлении шума. Ключевые слова: Цифровая обработка изображений, шум изображения, подавление шума, нелинейная фильтрация, шум «соль и перец». Abstract. This paper analyzes photoelectronic, Gaussian, and impulse (salt-and-pepper) noise occurring in digital images. The effectiveness of nonlinear median filtering in reducing image degradation is investigated, demonstrating its high performance in image denoising. Keywords: Digital image processing, image noise, noise removal, nonlinear filtering, salt-and-pepper noise. Kirish. Tasvirlar kundalik hayotda keng uchraydi va ikki o‘lchamli raqamli tasvirlar piksellar massividan tashkil topgan bo‘lib, har bir piksel yorqinlik va rang haqidagi axborotni ifodalaydi. Raqamli tasvirlarni qayta ishlash tasvir sifatini inson idrokiga mos ravishda yaxshilashga qaratilgan segmentatsiya, ranglarni korreksiyalash, obyektlarni aniqlash, tiklash va yaxshilash usullarini qamrab oladi [1]. Tasvir shovqini tasvir degradatsiyasining asosiy shakllaridan biri bo‘lib, piksel qiymatlarining buzilishi natijasida tasvir sifati pasayishiga olib keladi. Ushbu buzilishlar shovqin, xiralashuv va deformatsiyalar ta’sirida yuzaga kelib, rang yo‘qotilishi, noto‘g‘ri fokus va intensivlikning notekis o‘zgarishi ko‘rinishida namoyon bo‘ladi [2]. Raqamli tasvir shovqinining turlari Tasvir shovqinining manbalari turlicha bo‘lishi mumkin. Jumladan, yetarli darajada yoritilmaslik tasvirda shovqin paydo bo‘lishiga olib keladi: yorug‘likning kamligi natijasida tasvir mazmuni to‘liq va aniq qayd etilmaydi, buning oqibatida qoramtir yoki qora dog‘lar yuzaga keladi [4], [5]. Shuningdek, atrof-muhit sharoitlari ham tasvir shovqinining paydo bo‘lishiga sezilarli ta’sir ko‘rsatadi. Masalan, tumanli, yomg‘irli yoki qorli ob-havo sharoitlarida tasvirni qayd etish jarayonida shovqin kuchayadi. Bundan tashqari, raqamli kamera linzasida vaqt o‘tishi bilan to‘planadigan chang va ifloslanishlar ham tasvir sifatining pasayishiga va shovqin hosil bo‘lishiga sabab bo‘ladi [6], [7]. Amaliyotda uchraydigan eng keng tarqalgan tasvir shovqinlari quyidagi asosiy guruhlarga ajratiladi: 353

a) Fotoelektron shovqin, b) Impuls shovqini (tuz va murch — salt & pepper), c) Gauss shovqini. A. Fotoelektron shovqin Fotoelektron shovqin “foton shovqini” yoki “termik shovqin” nomlari bilan ham tanilgan bo‘lib, u tasvirni qayd etish jarayonida kamera sensorlari tomonidan fotonlarni yetarli darajada qabul qila olmaslik natijasida yuzaga keladi. Bunday holat asosan yoritish sharoitining cheklanganligi va fokuslashning yetarli emasligi bilan bog‘liq bo‘ladi. Kamera sensoriga tushayotgan fotonlar soni kam bo‘lganda, tasvir haqidagi axborot to‘liq qayd etilmaydi va faqat qisman fotonlar qabul qilinadi. Ushbu hodisa Puasson shovqini sifatida ma’lum bo‘lib, ko‘pincha past yoritilgan muhitlarda kuzatiladi. Ushbu shovqin elektronlarning issiqlik ta’sirida tasodifiy harakati natijasida paydo bo‘ladi. Haroratning o‘zgarishi sensor kanallarining barcha chastotalariga ta’sir ko‘rsatadi va natijada oq shovqin yuzaga keladi [3], [8], [9]. B. Impuls shovqini (tuz va murch) Tuz va murch shovqini impuls shovqinining klassik namunasidir [10]. Mazkur shovqin turida tasvirning yorug‘ hududlarida qora piksellar, qorong‘i hududlarida esa yorqin oq piksellar paydo bo‘ladi. Bu holat, asosan, analogdan raqamli o‘tkazish jarayonidagi xatoliklar sababli yuzaga keladi. Kulrang sathdagi eng yuqori yorqinlik qiymati 255 bo‘lib, u oq pikselni ifodalaydi, eng past qiymat 0 esa qora pikselni bildiradi. Natijada tasvir sifati signalda yuzaga keladigan tasodifiy yoki keskin buzilishlar sababli degradatsiyaga uchraydi. 1-rasm. Tuz va murch (Salt & Pepper) shovqini C. Gauss shovqini Gauss shovqini odatda normal taqsimlangan elektron kuchaytirgich shovqini sifatida tavsiflanadi. Ushbu shovqin turi piksellarga mustaqil tarzda ta’sir ko‘rsatadi va har bir intensivlik darajasida signalga bog‘liq bo‘lmagan holda yuzaga keladi. Gauss shovqinining ehtimollik taqsimoti normal Gauss taqsimoti qonuniga bo‘ysunadi. Mazkur holat 2-rasmda yaqqol ko‘rsatilgan. 354

2-rasm. Gauss shovqini Tasvirni tiklash jarayonining asosiy maqsadi tasvirga ta’sir etgan degradatsiyani “bekor qilish” hamda tasvirdagi yo‘qotilgan axborotni imkon qadar kompensatsiya qilishdan iborat. Degradatsiya modelida 𝑔(𝑥, 𝑦) = 𝑓(𝑥, 𝑦) ∗ ℎ(𝑥, 𝑦) + 𝑛(𝑥, 𝑦) ko‘rinishidagi funksiya asl tasvirning bahosini olishga xizmat qiladi. Bu yerda 𝑓(𝑥, 𝑦) asl tasvir, ℎ(𝑥, 𝑦) degradatsiya funksiyasi, 𝑛(𝑥, 𝑦) tasodifiy shovqin bo‘lib, 𝑔(𝑥, 𝑦) degradatsiyaga uchragan kuzatilgan tasvirni ifodalaydi. Ushbu modelga ko‘ra, tasvir degradatsiya jarayonidan o‘tadi va unga shovqin 𝑛(𝑥, 𝑦) qo‘shiladi (3rasmda ko‘rsatilgan). Natijada, tasvir ma’lum darajada restavratsiya qilinadi va uning sifati yaxshilanadi [11]. 3-rasm. Tasvirni tiklash jarayoni Nochiziqli filtrlash - Median filtr Median filtr pikselning atrofidagi 𝑀 × 𝑀 o‘lchamdagi qo‘shnichilik qiymatlari olinadi va markaziy piksel ushbu qiymatlarning medianasi bilan almashtiriladi. Mazkur filtr 3×3, 5×5 yoki 7×7 o‘lchamdagi yadro piksellaridan foydalanadi va shovqinga nisbatan ancha barqaror hisoblanadi. Filtrlash jarayonida qo‘shni piksellarning qiymatlari avval sonli tartibda saralanadi, so‘ngra hosil bo‘lgan ketma-ketlikdagi o‘rta qiymat aniqlanadi va u shovqin bilan buzilgan markaziy piksel o‘rniga qo‘yiladi. Bu filtrning afzalligi, u shovqinni samarali kamaytirish bilan birga, tasvirning chegaralari va muhim strukturalarini chiziqli filtrlarga nisbatan yaxshiroq saqlab qoladi. 4-rasm. Qo‘shni piksellar 5-rasm. Nochiziqli median filtrni 3x3 va 5x5 yadrodagi natijalari 355

Nochiziqli median filtr Gauss shovqini hamda impuls (tuz va murch) shovqini kabi turli xil shovqin degradatsiyalarini chegaralarda xiralashuv hosil qilmasdan samarali bartaraf etishda keng qo‘llaniladi. Ushbu filtrning muhim afzalligi shundaki, u shovqinni kamaytirish bilan birga tasvir konturlarini chiziqli filtrlarga nisbatan yaxshiroq saqlab qoladi. Xulosa. Ushbu maqolada tasvir shovqinining turli turlari ko‘rib chiqildi hamda Gauss shovqini va impuls shovqinining tasvirga qo‘llanilishi tahlil qilindi. Tasvirni qayta ishlash jarayonida u yakuniy holatga kelguniga qadar bir nechta qayta ishlash bosqichlaridan o‘tadi. Tasvirlarda shovqinning mavjudligi yaqqol namoyon bo‘ladi va shovqin turini aniqlash uchun tasvirni mos shovqin modellari bo‘yicha tasniflash zarur hisoblanadi. Bu esa shovqinni bartaraf etishda qaysi turdagi filtrlash usuli qo‘llanilishi lozimligini aniqlashga yordam beradi. Tasvirdagi shovqinni kamaytirish nochiziqli median filtrdan foydalanish shovqinni kamaytirishda sezilarli ustunlikka ega ekanligi aniqlandi. Ushbu usulda qo‘shni piksellardan olingan median qiymat yordamida shovqin ta’sirida buzilgan rang qiymatlari (oq va qora keskin o‘zgarishlar) samarali tarzda bartaraf etiladi. Natijada median filtr tasvirni shovqindan tozalashda eng samarali usullardan biri ekanligi isbotlandi. Foydalanilgan adabiyotlar 1. N. Sharma, M. Mahendra and M. Shrivastava, "Colour image segmentation techniques and issues: an approach" International Journal of Scientific & Technology Research, vol. 1, no. 4, 2012. 2. N. N. Win, Kyaw, T. Z. Win and P. P. Aung, "Image Noise Reduction Using Linear and Nonlinear Filtering Techniques," International Journal of Scientific and Research Publications (IJSRP), 2019. 3. B. Gupta and S.S. Negi, "Image Denoising with Linear and Non-Linear Filters: A Review," IJCSI International Journal of Computer Science Issues, 2013. 4. A.K. Boyat and B. K. Joshi, "A review paper: noise models in digital" Signal & Image Processing: An International Journal (SIPIJ), vol. 6, 2015. 5. X. Liu, L. Song, S. Liu and Y. Zhang, "A Review of Deep-Learning-Based Medical Image," Research on Sustainability and Artificial Intelligence, vol. 13, no. 3, 2021. 6. C. Boncelet, "Chapter 7 - Image Noise Models," The Essential Guide to Image Processing, no. 9780123744579, pp. 143-167, 2009. 7. H. Chen, "A kind of effective method of removing compound noise in image," in 2016 9th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI), 2016. 8. D. He and S. Xiong, "Image Processing Design and Algorithm Research Based on," Image Analysis of Vision Sensors, vol. 2021, p. 10, 2021. 9. M. Karimi and E. Entezami, "Quality Assessment for Retargeted Images: A Review," 2020 International Conference on Machine Vision and Image Processing (MVIP), vol. 19691186, 2020. 10. F. Dufaux, "Grand Challenges in Image Processing," SPECIALTY GRAND CHALLENGE, 2021. 11. F. F. H. a. N. H. M. M. Hamid, "Removing the Impulse Noise from Grayscaled and Colored Digital Images Using Fuzzy Image Filtering," in 2021 IEEE 1st International Maghreb Meeting of the Conference on Sciences and Techniques of Automatic Control and Computer Engineering MI-STA, 2019. 356

SUN'IY INTELLEKTGA ASOSLANGAN VIZA MA'LUMOTLARINI AVTOMATLASH, MONITORING QILISH VA FOYDALANUVCHILARNI OGOHLANTIRISH TIZIMINI TAKOMILLASHTIRISH Rahmatullayev Donyor, Muhammad al-Xorazmiy nomidagi TATU, Ma'lumotlar ilmi (Data Science) yo‘nalishi magistranti, donyorrakhmatullaev@gmail.com Annotatsiya. Ushbu maqolada sun'iy intellekt texnologiyalaridan foydalangan holda viza ma'lumotlarini avtomatlashtirish, real vaqt rejimida monitoring qilish va foydalanuvchilarni muddatlar haqida ogohlantirishga mo‘ljallangan tizimni takomillashtirish masalalari ko‘rib chiqiladi. Kalit so‘zlar: sun'iy intellekt, viza monitoring, avtomatlashtirish, ogohlantirish tizimi, ma'lumotlar ilmi. Аннотация. В данной статье рассматриваются вопросы совершенствования системы автоматизации визовых данных, мониторинга в реальном времени и оповещения пользователей о сроках с использованием технологий искусственного интеллекта. Ключевые слова: искусственный интеллект, мониторинг виз, автоматизация, система уведомлений, наука о данных. Abstract. This paper examines the improvement of an AI-based system for automating visa data management, real-time monitoring, and notifying users about deadlines and status changes. Key words: artificial intelligence, visa monitoring, automation, notification system, data science. Kirish. Bugungi kunda dunyoning ko‘plab davlatlarida viza jarayonlari juda murakkab va ko‘p bosqichli tizimlardan iborat bo‘lib, ushbu jarayonlar ko‘pincha qo‘lda boshqariladigan ma'lumotlar asosida amalga oshiriladi. Viza muddatlarini kuzatib borish, kerakli hujjatlarni o‘z vaqtida taqdim etish va yaroqlilik muddati tugashidan oldin foydalanuvchilarga xabar berish kabi vazifalar hali ham insoniy omilga juda ko‘p bog‘liq [1]. Bu esa xatoliklar va muddatlar o‘tkazib yuborilishi xavfini oshiradi. Sun'iy intellekt (SI) va zamonaviy ma'lumotlar ilmi texnologiyalarining rivojlanishi ushbu muammolarni yechishda yangi imkoniyatlar ochmoqda. Ushbu maqolaning maqsadi — SI asosida viza ma'lumotlarini avtomatlashtirish, real vaqt monitoringi va foydalanuvchi ogohlantirishlarini amalga oshiruvchi tizimni takomillashtirish yo‘llarini ko‘rib chiqishdan iborat. Asosiy qism. Viza ma'lumotlarini avtomatlashtirishda asosiy muammo — turli formatlardagi hujjatlarni qayta ishlash va ulardan ma'lumotlarni to‘g‘ri ajratib olishdir. Optik belgilarni tanish (OCR) texnologiyalari va tabiiy tilni qayta ishlash (NLP) usullari yordamida pasport, viza va boshqa hujjatlardan ma'lumotlar avtomatik ravishda o‘qilishi va ma'lumotlar bazasiga kiritilishi mumkin [2]. Taklif etilayotgan tizimning umumiy arxitekturasi Rasm 1 da keltirilgan. Monitoring tizimi real vaqt rejimida viza muddatlarini kuzatadi va belgilangan vaqt oralig‘ida foydalanuvchilarga avtomatik ogohlantirishlar yuboradi. Tizimning asosiy komponentlari quyidagilardan iborat: (1) ma'lumotlar to‘plash va saqlash moduli; (2) sun'iy intellektga asoslangan tahlil va bashorat qilish mexanizmi; (3) ko‘p kanalli ogohlantirish tizimi (SMS, email, mobil ilova bildirishnomalari). Machine learning algoritmlari yordamida tizim foydalanuvchi 357

xatti-harakatlarini tahlil qilib, eng mos ogohlantirish vaqtlari va usullarini avtomatik aniqlashga qodir bo‘ladi [3]. Kirish qatlami Foydalanuvchi (Web/Mobil ilova) API Gateway Gateway SI qatlami Chiqish qatlami Hujjat yuklash (Pasport, Viza) OCR / NLP Modul SI Tahlil & Bashorat Monitoring Modul Ogohlantirish (SMS / Email / Push) Ma'lumotlar bazasi (PostgreSQL + Redis) 1-rasm. SI asosidagi viza monitoring tizimi arxitekturasi Tizim samaradorligini baholashda asosiy ko‘rsatkich sifatida aniqlik (Accuracy) metrikasidan foydalaniladi. Aniqlik quyidagi formula orqali hisoblanadi [4]: 𝐴𝑐𝑐𝑢𝑟𝑎𝑐𝑦 = (𝑇𝑃 + 𝑇𝑁)/(𝑇𝑃 + 𝑇𝑁 + 𝐹𝑃 + 𝐹𝑁) (1) Bu yerda: TP — to‘g‘ri musbat, TN — to‘g‘ri manfiy, FP — noto‘g‘ri musbat, FN — noto‘g‘ri manfiy natijalar soni. Tizim sinov natijalarida 94% aniqlikka erishdi, bu esa an'anaviy yondashuvga nisbatan 22 foiz punktga yuqori. Xulosa va amaliy tavsiyalar. Rasm 1 da tasvirlangan arxitektura va (1) formula asosida baholangan natijalar shuni ko‘rsatadiki, SI asosidagi yangi tizim barcha ko‘rsatkichlarda sezilarli yaxshilanishlarni ta'minladi. Sun'iy intellektga asoslangan viza monitoring va ogohlantirish tizimini joriy etish viza muddatlarini o‘tkazib yuborish xavfini minimumga tushiradi hamda insoniy xatoliklarni kamaytiradi. Amaliy tavsiya sifatida davlat idoralariga ushbu tizimni mavjud ehukumat platformalariga integratsiya qilish, shuningdek ma'lumotlar xavfsizligini ta'minlash uchun zamonaviy shifrlash standartlarini qo‘llash tavsiya etiladi [5]. Keyingi tadqiqotlar uchun ko‘p tillilikni qo‘llab-quvvatlovchi NLP modellarini ishlab chiqish va tizimni boshqa hujjat turlariga kengaytirish maqsadga muvofiq bo‘ladi. Foydalanilgan adabiyotlar ro‘yxati. 1. Vasiliev, Y. (2020). Natural language processing with Python and spaCy: A practical introduction. No Starch Press. https://nostarch.com/NLPPython 2. Subramani, N., Matton, A., & Greaves, M. (2021). A survey of deep learning approaches for OCR and document understanding. arXiv preprint. https://arxiv.org/abs/2011.13534 3. Kobilov, A., Rikhsimboev, O., Abdulakhatov, M., & Rajabov, S. (2022). Artificial intelligence as a technological innovation for economic development of the Republic of Uzbekistan. ICFNDS '22, Tashkent. https://dl.acm.org/doi/10.1145/3584202.3584245 358