Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Ahmet Tezcan Tekin has a total of 20 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Uncertainty-Aware Reconciliation of Conflicting Sources on the Probability Simplex via Choquet Aggregation over Fuzzy Measures

    by Ahmet Tezcan Tekin

    We study the reconciliation of several biased, partially-overlapping estimators of a latent distribution on the probability simplex (the set of non-negative credit shares summing to one) when no ground truth is available to adjudicate their disagreement—the situation faced when conversion credit is reported by independent commercial attribution sources. We cast this as non-additive information fusion and develop a four-layer fuzzy framework: a reliability-weighted fuzzy representation, an interaction-corrected conflict measure, a Choquet integral with respect to an elicited fuzzy measure (capacity) followed by an L1-closure onto the simplex (a per-cell renormalization returning a valid distribution), and a Mamdani layer producing a conformally calibrated uncertainty band. Our theoretical contributions are (i) a simplex-consistency result—channel-wise Choquet aggregation with L1-closure always returns a valid distribution and reduces to the weighted mean under an additive capacity; (ii) variational and axiomatic characterizations of the operator; (iii) an exact Shapley–disagreement decomposition that quantifies how non-additivity reweights redundant sources; and (iv) a Lipschitz stability bound that guarantees robustness to noisy inputs. The capacity is elicited from observable signals alone, so the method requires no labels. On a firewalled, ten-seed benchmark across eight regimes the framework attains the lowest MAE in five of eight conditions and the lowest KL in six, improving on Dempster–Shafer, correlation-clustering, and reliability-weighted baselines by 1.6–2.3% (MAE) and 3.1–4.4% (KL)—significant in four of its five targeted redundancy- and bias-dominated regimes—with the conformal step attaining nominal coverage. This advantage stems from the elicited capacity rather than the Choquet operator (a simpler ordinal integral matches it on point error); Choquet is retained for its cardinal, calibratable output. An illustrative two-source real-data study exposes a 13.7% cost-per-install distortion in the single-source view, with the (uncalibrated) band widest where the sources disagree most.


  • Bulanık Eşleşme Yöntemi ile Veri Setlerindeki Hatalı Girişlerin Düzeltilmesi ve Makine Öğrenmesi Performansına Etkisi

    by Ahmet Tezcan Tekin

    Veri kalitesi, makine öğrenmesi modellerinin başarısını doğrudan etkileyen en kritik faktörlerden biridir. Gerçek dünya veri setlerinde sıklıkla karşılaşılan yazım hataları, tutarsız kodlamalar ve standart dışı girişler, model eğitim sürecini olumsuz yönde etkilemektedir. Bu çalışma, bulanık eşleşme (fuzzy matching) yöntemlerinin veri ön işleme aşamasında hatalı girişlerin tespiti ve düzeltilmesindeki rolünü incelemektedir. Levenshtein mesafesi, Jaro-Winkler benzerliği ve n-gram tabanlı yaklaşımlar gibi temel bulanık eşleşme algoritmalarının matematiksel temelleri sistematik biçimde ele alınmış ve bu yöntemlerin kategorik değişkenlerdeki tutarsızlıkları gidermede nasıl kullanılabileceği tartışılmıştır. Çalışmada, 10.000 kayıttan oluşan sentetik bir müşteri segmentasyonu veri setine kontrollü biçimde yazım hataları, büyük/küçük harf tutarsızlıkları, kısaltma farklılıkları ve transpozisyon hataları enjekte edilmiştir. Jaro-Winkler benzerliğine dayalı üç aşamalı bir düzeltme pipeline'ı uygulanarak hatalı girişlerin %96,3'ü başarılı biçimde düzeltilmiştir. Düzeltilmiş ve düzeltilmemiş veri setleri üzerinde Lojistik Regresyon, Karar Ağacı, Rastgele Orman ve XGBoost gibi sınıflandırma algoritmalarının performansları karşılaştırmalı olarak değerlendirilmiştir. Deneysel sonuçlar, bulanık eşleşme tabanlı veri düzeltmenin tüm modellerde F1 skorunu ortalama +8,9 yüzde puan artırdığını ve düzeltilmiş veri setinin orijinal hatasız veriye çok yakın performans ürettiğini ortaya koymuştur. Ayrıca düzeltme işleminin one-hot encoding sonrası öznitelik uzayını %38 oranında daraltarak hesaplama verimliliğini de önemli ölçüde iyileştirdiği gözlemlenmiştir.

  • Open Access

    Fuzzy Rule-Based Explanations for Tabular Black-Box Classifiers: A Comprehensive Empirical Framework with Prediction-Boundary-Aware Partitioning and Rule-Level Uncertainty Indication

    by Ahmet Tezcan Tekin

    Existing post hoc XAI (Explainable Artificial Intelligence) methods produce numerical attributions without symbolic structure (SHAP, LIME), low-coverage local rules (Anchors), or crisp tree surrogates without an interpretable rule-level uncertainty proxy. We present a fuzzy rule-based explanation framework for tabular black-box classifiers, extracting global IF–THEN rules with linguistic labels. This was validated on a 13-dataset benchmark with four model families (Wilcoxon, Friedman, TOST equivalence): (i) prediction-boundary-aware fuzzy partitioning raises mean fidelity from a vanilla Wang–Mendel baseline of 0.736 to 0.893 (+10.4 pp excluding the Breast Cancer outlier; +15.7 pp aggregate, both transparently reported); (ii) fired-rule consequent entropy provides a zero-cost rule-level uncertainty proxy (Spearman ρ = 0.420 with model prediction entropy, significant on 11/12 datasets—moderate by Cohen’s convention, with a 4/12 weak-correlation tail; complementary to probability-entropy and margin baselines). Fidelity is statistically equivalent to tree surrogates on classification (TOST p = 0.002, δ = 0.05) at ≈100% coverage. SHAP/LIME are excluded from the formal stability ranking because the perturbation metric measures the wrapped black-box rather than the attribution vector; cross-explainer comparison is reported in grouped form (full-coverage surrogates vs. local-coverage methods). On continuous regression (California Housing fidelity 0.422 vs. TreeSurrogate 0.840) and XOR-type multi-feature interactions, the framework is structurally weaker, addressed by a planned TSK extension.


  • Open Access

    Fuzzy Graded Preprocessing for Robust Machine Learning: A Three-Stage Mamdani Framework with Interpretable Audit Trails

    by Ahmet Tezcan Tekin

    Data preprocessing methods for machine learning overwhelmingly rely on binary logic—a value is either valid or invalid—and the corrective action does not scale with error severity. This paper introduces GDEDC, a Mamdani-type fuzzy inference framework that replaces binary preprocessing with graded error detection and proportional correction. Operating in three stages—fuzzy anomaly scoring, nine-rule Mamdani FIS classification, and sigmoid-weighted imputation—the framework corrects each value in proportion to its estimated error severity while retaining 100% of observations and producing a human-readable audit trail. We evaluate GDEDC on five UCI datasets and the Pima Indians diabetes dataset with five classifiers across six noise levels (5–30%), comparing against five baselines including MICE. Under leakage-free conditions, deletion-based methods consistently underperform raw data, while correction-based methods (GDEDC, KNN Imputation, MICE) deliver significant improvements. GDEDC matches KNN Imputation and MICE at low noise and surpasses both at ≥20% noise: on noise-sensitive classifiers, GDEDC achieves the best Friedman rank at 20–30% noise. Real-world validation on the Pima dataset confirms generalizability, with GDEDC outperforming IQR by +2.97% (p < 0.001, d = 0.684). Ablation analysis shows that sigmoid-based proportional correction is the primary contributor (+2.02 pp), and the full pipeline outperforms every ablated variant at 10–20% noise.


  • Customer Lifetime Value Prediction in Mobile Gaming Industry: Fuzzy Logic Approach

    by Ahmet Tezcan Tekin

    The Customer Lifetime Value (CLV) is an essential metric in customer relationship management (CRM), allowing companies to identify valuable customers and refine their advertising strategies. Traditional customer lifetime value prediction methods, including regression and machine learning techniques, frequently depend on accurate and predictable input data, making them less effective at capturing the inherent uncertainty and unpredictability in customer behavior. This research presents a fuzzy logic-based Customer Lifetime Value prediction model that integrates Recency, Frequency, and Monetary Value (RFM) as essential input factors. The proposed approach utilizes fuzzy membership functions and fuzzy inference systems (FIS), enabling consumers to possess partial membership in different CLV categories, hence offering a more adaptable and comprehensible framework for CLV calculation. A rule-based IF-THEN fuzzy system is established to categorize clients into various CLV segments, and defuzzification methods are employed to derive a precise CLV score. Experimental results indicate that the fuzzy logic model adeptly manages uncertainty and imprecision, outperforming traditional hard-segmentation methods by providing a continuous and adaptable strategy for CLV prediction. This research underscores the benefits of fuzzy logic in customer analytics, offering enterprises an easy and flexible instrument for customer segmentation, retention strategies, and revenue optimization.


  • Defect Detection in Manufacturing via Machine Learning Algorithms

    by Enes Şanlıtürk, Ahmet Tezcan Tekin, Ferhan Çebi

    In today's highly competitive global market, defects in manufacturing cause additional costs and these costs decrease the profit margin. So, defect-free products are crucial for firms. There are various studies conducted with different methodologies in different fields related to defect detection in the literature. Although one of these methodologies is machine learning (ML) algorithms, the literature review shows very few studies using these algorithms to detect physical product defects. This study aims to utilize machine learning algorithms for predicting defects in the manufacturing. Different ML algorithms are used to predict defected products. Then the performances of the algorithms are compared to find the most appropriate algorithm. The study is conducted on one of the leading manufacturing companies in Turkey, and the data is provided from the real-life problem of the company. With the proposed algorithm, it is desirable to predict defected products that may occur during the powder coating phase.


  • Stock Price Prediction

    by Ahmet Tezcan Tekin, Ferhan Çebi, Tolga Kaya

    In this study, the last two years' hourly opening and closing prices of the banks' stocks traded on BIST-30 were used as the dataset. The research is aimed to predict the closing prices of these stocks in the light of machine learning. In this context, the authors propose a new method containing ensemble learning algorithms and fuzzy clustering technics for predicting stock prices. With this method, authors aim to find stocks which are similar characteristics with test sets and model them together. Thanks to this method, authors aim to improve modelling success. For comparing the results, authors also create models with classical machine learning methods such as support vector machines, random forest, and boosting type new generation algorithms such as extreme gradient boosting and catboost.