Browsing by Author "Cakar, Tuna"

Now showing 1 - 8 of 8

Citation - Scopus: 2
AI-Driven Digital Soil Mapping: Leveraging Generative AI, Ensemble Learning and Geospatial Technologies for Data-Scarce Regions
(Springer Science and Business Media Deutschland GmbH, 2025) Drias, Yassine; Drias, Habiba; Belkadi, Widad Hassina; Cakar, Tuna; Abdelhamid, Zakaria; Bensemmane, Riad Yacine
This study presents a methodology for generating highresolution digital soil maps by integrating artificial intelligence (AI) with geospatial technologies. The research begins with comprehensive data collection and the extraction of relevant soil properties with the help of generative AI. To improve predictive accuracy, ensemble learning algorithms were employed due to their ability to capture complex relationships within soil characteristics. Additionally, a structured preprocessing pipeline was developed to refine and standardize the collected soil data, ensuring its suitability for modeling. The model's performance was evaluated using spatial cross-validation techniques to identify the most effective predictive approach. To validate the proposed methodology, experiments were conducted in northern Algeria, a region characterized by diverse landscapes ranging from arid zones to fertile plains. The results demonstrate the potential of AI-driven approaches to enhance soil mapping, particularly in regions where high-quality and up-to-date soil data are limited. This study underscores the efficacy of AI and geospatial technologies in advancing precision agriculture and land management.
Customer Segmentation and Churn Prediction via Customer Metrics
(IEEE, 2022) Bozkan, Tunahan; Cakar, Tuna; Sayar, Alperen; Ertugrul, Seyit
In this study, it is aimed to predict whether customers operating in the factoring sector will continue to trade in the next three months after the last transaction date, using data-driven machine learning models, based on their past transaction movements and their risk, limit and company data. As a result of the models established, Loss Analysis (Churn) of two different customer groups (Real and Legal factory) was carried out. It was estimated by the XGBoost model with an F1 Score of 74% and 77%. Thanks to this modeling, it was aimed to increase the retention rate of customers through special promotions and campaigns to be made to these customer groups, together with the prediction of the customers who will leave. Thanks to the increase in retention rates, a direct contribution to the transaction volume on a company basis was ensured.
Ensemble-Based Stock Prediction for Retail - XGBoost and LightGBM with Rolling Window Training
(Institute of Electrical and Electronics Engineers Inc., 2025) Patel, Jay Nimish; Kizilay, Ayse; Sahin, Zeynep; Sercan, Busra; Toprak, Samet; Cakar, Tuna
Stock prediction in retail settings is a critical challenge that impacts numerous businesses globally, that require precise and timely forecasts to optimize inventory management and enhance customer satisfaction. State-of-the-art approaches for accurate stock prediction leverage machine learning (ML) models, which require large amounts of historical sales data for effective training. Such detailed datasets are often hard to obtain, limiting the performance and scalability of these approaches. In this paper, we propose various strategies to tackle this limitation. Initially, we adopt a transfer-learning approach, utilizing pre-trained models like XGBoost and LightGBM, which are fine-tuned for stock prediction in retail environments. To further boost model performance, we incorporate an ensemble method that combines predictions from both models to improve accuracy and manage outliers. Experiments conducted on an extremely large dataset, comprising millions of retail transactions, highlight the presence of significant outliers. Our models, augmented with ensemble strategies, significantly outperform traditional models in handling these complexities and improving prediction accuracy. © 2025 Elsevier B.V., All rights reserved.
Financial Inputs Prediction with Machine Learning and Covariance Matrix Applications
(Institute of Electrical and Electronics Engineers Inc., 2025) Benli, Harun; Gunes, Peri; Ulkgun, Mert; Cakar, Tuna
Reliable estimation of the time-varying covariance matrix of asset returns is indispensable for portfolio construction, risk budgeting, and automated advisory services. Conventional estimators-rolling-window sample covariances, EWMA filters, and GARCH families-remain anchored to historical prices and therefore adapt slowly when market conditions pivot. To overcome this latency, we propose an end-to-end, machine-learning-driven framework that forecasts future covariances directly from high-frequency equity data, largely decoupling risk estimation from past observations. Our pipeline ingests heterogeneous stock feeds through a real-time API, applies error-minimising imputation (forward/backward fill, spline, VAR, wavelet, and co-kriging), and standardises returns via empirically selected scaling schemes. The processed features are then fed to a model zoo comprising linear and penalised regressions, tree ensembles (Random Forest, XGBoost, LightGBM, CatBoost), and kernel-based Support Vector Regression. Weekly walk-forward evaluation on a universe of Turkish insurance equities shows that LightGBM and SVR cut out-of-sample covariance prediction error by up to 35 % versus classical benchmarks. We embed the predicted matrices into five allocation engines-Markowitz mean-variance, Black-Litterman, minimum-variance, Risk Parity, and CVaR optimisation-demonstrating that Risk Parity delivers the most consistent variance reduction across 15 stock pairs, while Black-Litterman excels for idiosyncratic combinations such as ANSGR-AKGRT. A granular analysis reveals that day-to-day sign changes in returns create structural breaks that generic regressors miss; augmenting the architecture with a volatility-state classifier and regime-specific learners markedly sharpens turning-point detection. Beyond statistical gains, the framework is production-ready: it is fully implemented in Python, runs on cloud notebooks, and plugs into robo-advisory dashboards. The study thus bridges academic advances in covariance prediction with operational portfolio management, paving the way for broader cross-sector deployment and future research on deep sequential models, transaction-cost awareness, and multi-asset scalability. © 2025 Elsevier B.V., All rights reserved.
Havayolu Reklamlarında İzleyicilerin Duygu Ölçümü: Microsoft Azure Face API ile Yüz Kodlama Uygulaması
(2025) Cakar, Tuna; Zengin, Burhanettin; Akpur, Akın
Amaç: Havayolu işletmeleri, müşterilerini uçuşlara çekmek ve marka imajlarını güçlendirmek için sık sık reklam kampanyaları düzenlerler. Ancak, reklam kampanyalarının etkinliği geleneksel yöntemlerle ölçüldüğünde sınırlı kalabilmektedir. Bu çalışmanın amacı yüz kodlama teknolojisinin havayolu reklamlarında duygu ölçüm performansının değerlendirilmesidir. Yöntem: Bu çalışma deneysel bir tasarıma sahip olup laboratuvar ortamında 40 denek ile yapılmıştır. Microsoft Azure Face API uygulamasından elde edilen sayısal duygu verileri istatistik paket program aracılığı ile analiz edilmiştir. Bulgular: Araştırmada final sahnesi olan reklamlar anlamlı şekilde final sahnesi olmayan reklama göre daha fazla mutluluk duygusu oluşturduğu tespit edilmiştir. Sonuç: Bu çalışma kapsamında kullanılan yüz kodlama yazılımının elde edilen çıktılar çerçevesinde ölçümlemede başarılı olduğu görülmüştür.
İnternet Trafik Hızının Tahmininde Derin Öğrenme ve Ağaç Tabanlı Modellerin Karşılaştırılması
(Institute of Electrical and Electronics Engineers Inc., 2025) Filiz, Gozde; Altintas, Suat; Yildiz, Aysenur; Kara, Erkan; Drias, Yassine; Cakar, Tuna
This study addresses the prediction of internet traffic speed using time-dependent data from an internet service provider through different modeling approaches. On an anonymized dataset, the performance of the moving average method, various deep learning models (N-BEATS, N-HITS, TimesNet, TSMixer, LSTM), and the XGBoost regression model enhanced with feature engineering was compared. Time series cross-validation and random hyperparameter search were used for model training. According to the results, the XGBoost model achieved the highest accuracy with 98.7% explained variance (R2), while among the deep learning models, N-BEATS and N-HITS achieved the best performance with R2 values around 90%. The findings indicate that tree-based methods supported by carefully selected features can offer higher accuracy and computational efficiency compared to complex deep learning models in internet traffic forecasting. © 2025 Elsevier B.V., All rights reserved.
Makine Öğrenimi ve Çok Boyutlu Anket Verileri Kullanılarak Öğrenci Başarısının Tahmini: Eğitim Programı Üzerine Bir Uygulama
(Institute of Electrical and Electronics Engineers Inc., 2025) Behsi, Zeynep; Dereli, Serhan; Cakar, Tuna; Patel, Jay; Cicek, Gultekin; Drias, Yassine
This study develops a machine learning model integrating survey data and performance metrics to predict student success in the UpSchool education program. Students' personality traits assessed by DISC analysis, financial management, social, and emotional skills were clustered into "Successful,""Unsuccessful,"and "Moderately Successful"groups using K-means clustering. The SMOTE technique addressed data imbalance issues, and algorithms such as Logistic Regression, Random Forest, LightGBM, and AdaBoost were tested. After hyperparameter optimization, AdaBoost and LightGBM achieved the highest predictive performance. Results demonstrated the effectiveness of machine learning models in forecasting student success in educational programs. Future studies are recommended to enhance model performance through expanded datasets and to validate the model's applicability across diverse educational contexts. © 2025 Elsevier B.V., All rights reserved.
Yapay Öğrenme Tabanlı Mikrofaktoring Skorlama Modeli ve Kredi Risk Yönetim Sistemi Geliştirilmesi
(Institute of Electrical and Electronics Engineers Inc., 2025) Sayar, Alperen; Ates, Yigit; Ertugrul, Seyit; Turan, Elif Naz; Drias, Yassine; Cakar, Tuna
Credit scoring systems are critical tools used by factoring institutions to assess the credit risks of SME businesses seeking microloans. This study presents a comprehensive predictive modeling framework that achieves 82.67% ROC-AUC with 65.34% Gini score on test data, demonstrating robust discriminative capability despite significant class imbalance. Our ensemble approach outperforms individual boosting models by leveraging their complementary strengths in payment behavior analysis and fraud detection. The raw data was cleaned, transformed, and optimized using the Polars library, with specialized features for detecting fraud patterns and time-based risk indicators. When implementing a score threshold of 950, our model significantly improves the detection of non-performing loans (NPL) compared to traditional rule-based approaches by reducing the net deficit from 6.59% to 2.62%. When applied to previously rejected applications, the model projects a potential 762.57% increase in transaction count and 747.05% growth in transaction volume. © 2025 Elsevier B.V., All rights reserved.