Browsing by Author "Arsan, T."

Now showing 1 - 6 of 6

Citation Count: 0
Building Damage Assessment To Facilitate Post-Earthquake Search and Rescue Missions by Leveraging a Machine Learning Algorithm
(Institute of Electrical and Electronics Engineers Inc., 2024) Zaker, M.; Alsan, H.F.; Arsan, T.
Earthquakes have a severe impact on people's lives and infrastructure. Many emergency institutes and search and rescue missions need accurate post-earthquake response strategies, particularly in building damage assessment. Traditional methods, relying on manual inspections, are inefficient compared to Machine Learning (ML) algorithms. Thus, Random Forest (RF) algorithms stand out because they handle diverse datasets effectively and minimize overfitting. The study outlines the methodology encompassing data preparation, exploratory analysis, feature engineering, and model building, employing a preprocessing pipeline integrating numerical and categorical features. Additionally, Principal Component Analysis (PCA) is applied to reduce dimensionality. The results of the RF model showed an accuracy of 94% and the highest F1-score of 97% among all the grades, demonstrating its efficacy in predicting damage grades post-earthquake. The results can help support better disaster management plans by helping to prioritize rescue operations and allocate resources wisely. © 2024 IEEE.
Citation Count: 0
Capacity Planning for Electricity Utility Call Centers: a Time Series Analysis Approach
(Institute of Electrical and Electronics Engineers Inc., 2024) Kavas, E.; Alsan, H.F.; Arsan, T.
Electric power systems are crucial for modern society, yet their reliability can be challenged by unforeseen disruptions, causing electricity supply disruptions. Call centers are essential for managing customer inquiries during such outages, acting as communication hubs for electricity utility companies. Effective capacity planning is vital for these call centers to maintain efficient operations and meet customer demands promptly. Proper workforce management ensures that enough skilled agents can handle calls effectively and maintain high service quality. Capacity planning begins with analyzing historical data to understand call volumes, patterns, and peak times. This data analysis identifies trends and factors influencing call patterns, enabling accurate forecasting of future demand and optimizing staffing levels. This paper provides a comprehensive overview of quantitative forecasting methods, focusing on Time Series Analysis applied to a dataset from a Turkish electric utility company that exhibits typical seasonal fluctuations. Specifically, the study examines the performance of AutoRegressive Integrated Moving Average and Seasonal AutoRegressive Integrated Moving Average models. Results indicate that both models perform well, with the Seasonal AutoRegressive Integrated Moving Average model demonstrating slightly superior performance compared to the AutoRegressive Integrated Moving Average model. This suggests that the Seasonal AutoRegressive Integrated Moving Average model may be more suitable for forecasting inbound calls at electricity utility call centers. This paper's detailed analysis and methodology offer valuable insights for optimizing operational efficiency, reducing costs, and enhancing customer satisfaction in dynamic and challenging operational scenarios. © 2024 IEEE.
Citation Count: 0
A Data Science Perspective on Global Trends in Energy Production
(Institute of Electrical and Electronics Engineers Inc., 2024) Hatira, N.; Alsan, H.F.; Arsan, T.
As global demand for energy continues to rise, understanding the trends and dynamics of energy generation is crucial to ensure a sustainable and efficient energy future. This study employs data science techniques to analyze global energy production data from 48 countries spanning 2010 to 2023. Initially, we use clustering methods to categorize countries based on their energy production profiles into three distinct groups: high, medium, and low production. This clustering provides insights into the diverse energy strategies and capacities across different regions. Subsequently, we apply and compare two classification models, specifically Random Forest and Gradient Boosting, to predict the dominant energy source for each cluster. Furthermore, we perform a comparative analysis of two forecasting models, SARIMA and Prophet, to predict future renewable energy production for countries with high production profiles, such as the USA and China. The forecasting results show the efficacy of these models in capturing seasonal trends and providing accurate predictions. © 2024 IEEE.
Citation Count: 1
Istanbul Dam Water Levels Forecasting Using ARIMA Models
(Institute of Electrical and Electronics Engineers Inc., 2022) Sekban, J.; Nabil, M.O.M.; Alsan, H.F.; Arsan, T.
River, sea, reservoir, and dam water levels are constantly measured by organizations and governmental bodies because of their environmental effects as well as their influence on human behavior. In this study, the monthly dam levels in Istanbul, Turkey, were predicted. Different models and configurations were compared to each other, and the best-performing model was identified. The models were based on conventional auto-regressive models (AR), moving average models (MA), auto-regressive moving average (ARMA), and ARMA with Exogenous variables (ARIMAX). © 2022 IEEE.
Citation Count: 1
Multimodal Retrieval With Contrastive Pretraining
(Institute of Electrical and Electronics Engineers Inc., 2021) Alsan, H.F.; Yildiz, E.; Safdil, E.B.; Arslan, F.; Arsan, T.
In this paper, we present multimodal data retrieval aided with contrastive pretraining. Our approach is to pretrain a contrastive network to assist in multimodal retrieval tasks. We work with multimodal data, which has image and caption (text) pairs. We present a dual encoder deep neural network with the image and text encoder to encode multimodal data (images and text) to represent vectors. These representation vectors are used for similarity-based retrieval. Image encoder is a 2D convolutional network, and text encoder is a recurrent neural network (Long-Short Term Memory). MS-COCO 2014 dataset has both images and captions, and it is used for multimodal training with triplet loss. We used a convolutional Siamese network to compute the similarities between images before the dual encoder training (contrastive pretraining). The advantage is that Siamese networks can aid the retrieval, and we seek to show if Siamese networks can be used in practice. Finally, we investigated the performance of Siamese assisted retrieval with BLEU score metric. We conclude that Siamese can help with image-to-text retrieval tasks. © 2021 IEEE.
Citation Count: 0
Multitype Learning Via Multimodal Data Embedding
(Institute of Electrical and Electronics Engineers Inc., 2021) Yildiz, E.; Safdil, E.B.; Arslan, F.; Alsan, H.F.; Arsan, T.
This paper creates a multimodal retrieval system for image and text data in a multi-type learning approach that enables text-to-image, image-to-text, text-to-text, and image-to-image retrievals. As a practical solution, a mobile application is developed in which the users can upload their images to search a description sentence for the images. The user system is created on the application, which is done with React Native, and crucial features like e-mail authentication and reset password options are added to the application. An essential database system is designed with PostgreSQL to store user information and search for the user. The multimodal embedding study is worked, and the model that recognizes multitype retrievals is formed. The image-to-text retrieval model, which is our application's idea, is applied to the mobile application. © 2021 IEEE.