Browsing by Author "Alsan,H.F."

Now showing 1 - 5 of 5

Citation Count: 0
Advancing Anomaly Detection in Time Series Data: A Knowledge Distillation Approach with LSTM Model
(Institute of Electrical and Electronics Engineers Inc., 2023) Arsan, Taner; Camlidere,B.; Yildiz,E.; Guler,A.K.; Alsan,H.F.; Arsan,T.
This paper focuses on enhancing anomaly detection in time series data using deep learning techniques. Particularly, it investigates the integration of knowledge distillation with LSTM-based models for improved precision, efficiency, and interpretability. The study outlines objectives such as dataset preprocessing, developing a novel LSTM-knowledge distillation framework, incorporating Grafana, InfluxDB, Flask API with Docker, performance assessment, and practical implications. Results highlight the efficacy of knowledge distillation in enhancing student model performance. The proposed approach enhances anomaly detection, offering a viable solution for real-world applications. © 2023 IEEE.
Citation Count: 0
Deciphering the Cluster-Specific Marker Genes via Integration of Single Cell RNA Sequencing Datasets
(Institute of Electrical and Electronics Engineers Inc., 2023) Arsan, Taner; Sogunmez,N.; Altaf,A.; Alsan,H.F.; Arsan,T.
Experimental data from brain tissues are critical for tackling the problems in brain development and revealing the underlying mechanisms of disease states. However, obtaining the brain tissue is a major challenge. Human brain organoids hold remarkable promise for this goal, but they suffer from substantial organoid-to-organoid variability. We performed a data-driven analysis on single-cell RNA-sequencing data using 17775 cells isolated from 2 individual organoids. The main goal was to accurately integrate the data coming from unmatched datasets, cluster the cells based on their similarity levels and predict the differentially expressed genes per cell types to reveal novel brain cell types and markers. This research opens a way to map human brain cells and develop novel and precise machine learning algorithms for accurate scRNA-Seq data analysis. © 2023 IEEE.
Citation Count: 0
Enhancing Robotic Performance: Analyzing Force and Torque Measurements for Predicting Execution Failures
(Institute of Electrical and Electronics Engineers Inc., 2023) Arsan, Taner; Alsan,H.F.; Arsan,T.
Robots play an important role in many sectors, automating processes and supplementing human talents. However, guaranteeing reliability is critical for effective integration and widespread adoption. As a result, forecasting and managing these errors is critical. This research examines force and torque measurements in order to better understand the causes and patterns of robot execution errors. We hope to build prediction models that improve robot design and performance, eventually boosting their reliability and efficacy, by using data analysis and machine learning approaches. This study's research aims include using a dataset of force and torque measurements to predict and define robot execution failures, We hope to uncover the complex links between force and torque measurements and failure types, find crucial signals or precursors to failures, and construct strong prediction models for correct failure categorization by tackling these research topics. This study contributes to data science by demonstrating the use of analytics approaches to improve the dependability and performance of robots in real-world scenarios. © 2023 IEEE.
Citation Count: 0
Machine Failure Prediction: : A Comparative Anomaly Detection
(Institute of Electrical and Electronics Engineers Inc., 2023) Arsan, Taner; Alsan,H.F.; Arsan,T.
Anomaly detection techniques seek to uncover unusual changes in the expected behavior of target indicators and, when used for intrusion detection, suspect assaults whenever the mentioned deviations are found. This technique is crucial in identifying and flagging abnormal instances in various domains. Several anomaly detection algorithms have been suggested, tested experimentally, and assessed in qualitative and quantitative surveys in the literature. However, there is a scarcity of comparative research, and methodological shortcomings are observed in existing studies. This paper investigates the performance of ten popular anomaly detection models for feature correlation analysis for predictive maintenance to detect machine failure with the most known approaches. The models considered are Local Outlier Factor (LOF), K-Nearest Neighbors (KNN), Support Vector Machines, Elliptic Envelope, Isolation Forest, Decision Tree, Extra Trees, Random Forest, AdaBoost, and Gradient Boosting. We evaluate the models using two scenarios: one with two correlated features and another with all features focused on correlated features. The evaluation metrics used for comparison are assessed by GridSearchCV and RandomizedSearchCV and compared to the cross-validation methods. © 2023 IEEE.
Citation Count: 0
Network Traffic Anomaly Detection Using Quantile Regression with Tolerance
(Institute of Electrical and Electronics Engineers Inc., 2023) Arsan, Taner; Guler,A.K.; Yildiz,E.; Kilinc,S.; Camlidere,B.; Arsan,T.
Network traffic anomaly detection describes a time series anomaly detection problem where a sudden increase or decrease (called spikes) in network traffic is predicted. Data is modeled with the trend and heteroscedastic noise component. Traditional autoregressive models struggle to capture data changes effectively, making anomaly detection difficult. Our approach is to generate upper and lower limits by using quantile regression. We use a deep learning based multilayer perceptron model to predict five data quantiles 1, 25, 50, 75, and 99. The upper and lower limits are calculated as differences between the quantile-1 and quantile-99. Any data that is outside these limits are considered as an anomaly. We also add tolerance to these limits to add flexibility to anomaly detection. Anomalies and non-anomalies are labeled to get a binary classification task. Anomaly detection is class imbalanced by nature; therefore, precision, recall, and F-1 score are computed to evaluate the proposed anomaly detection method. We conclude that choosing tolerance is a tradeoff between false alarms and missing anomaly detections. © 2023 IEEE.