Browsing by Author "Tokuc, A. Aylin"

Now showing 1 - 2 of 2

Citation - WoS: 4
Click Prediction Boosting Via Bayesian Hyperparameter Optimization-Based Ensemble Learning Pipelines
(Elsevier, 2023) Demirel, Cagatay; Tokuc, A. Aylin; Tekin, Ahmet Tezcan
Online travel agencies (OTA's) advertise their website offers on meta-search bidding engines. The problem of predicting the number of clicks a hotel would receive for a given bid amount is an important step in the management of an OTA's advertisement campaign on a meta-search engine because bid times number of clicks defines the cost to be generated. Various regressors are ensembled in this work to improve click prediction performance. After preprocessing, the entire feature set is divided into 5 groups, with the training set preceding the test set in the time domain, and multi-set validation is applied. The training data for each validation set is then subjected to feature elimination, and the selected models are next validated with separate ensemble models based on the mean and weighted average of the test predictions. Additionally, a stacked meta-regressor is designed and tested, along with the complete train set, whose click prediction values are extracted in accordance with the out- of-fold prediction principle. The original feature set and the stacked input data are then combined, and level-1 regressors are trained once again to form blended meta-regressors. All individually trained models are then compared pairwise with their ensemble variations. Adjusted R 2 score is chosen as the main evaluation metric. The meta-models with tree-based ensemble level-1 regressors do not provide any performance improvement over the stand-alone versions, whereas the stack and blended ensemble models with all other non-tree-based models as level-1 regressors boost click prediction (0.114 and 0.124) significantly compared to their stand-alone versions. Additionally, statistical evidence is provided to support the importance of Bayesian hyperparameter optimization to the performance-boosting of level-1 regressors.
Citation - WoS: 0
Citation - Scopus: 0
Predicting User Purchases From Clickstream Data: a Comparative Analysis of Clickstream Data Representations and Machine Learning Models
(IEEE-Inst Electrical Electronics Engineers inc, 2025) Dağ, Tamer; Dag, Tamer; Computer Engineering
Predicting purchase events from e-commerce clickstream data is a critical challenge with significant implications for optimizing marketing strategies and enhancing customer experience. This study addresses this challenge by systematically evaluating and comparing multiple data representations - aggregated session attributes, recent user actions, and hybrid combinations - which bridges gaps in the existing literature and demonstrates the superiority of hybrid approaches. Unlike prior research, which typically focuses on single representations, our approach combines aggregated session-level summaries with granular, sequential user actions to capture both long-term and short-term behavioral patterns. Through comprehensive experimentation, we compared multiple machine learning models, including LightGBM, decision trees, gradient boosting, SVC, and logistic regression, using real-world e-commerce clickstream data. Notably, the hybrid representation with LightGBM achieved superior predictive performance, significantly outperforming alternative methods. Feature importance analysis revealed key factors influencing purchase likelihood, such as time since the last event, session duration, and product interactions. This study provides actionable insights into real-time marketing interventions by demonstrating the practical utility of hybrid data representations and efficient tree-based models. Our findings offer a scalable and interpretable framework for e-commerce platforms to enhance purchase predictions and optimize marketing strategies.