| dc.description.abstract | The widespread use of smartphones by most of the population has made smartphone data highly valuable in gaining insights into users’ behavior by capturing their daily activities. This data can be leveraged to predict user behavior and intentions, forecast their location and suggest places and activities. Machine learning and deep learning techniques have been employed to extract hidden and valuable information from complex and diverse behavioral data to improve the ability to recognize and utilize human activities across different domains. The data is collected using various sensors, and learning algorithms generate valuable predictions in fields such as medicine, travel, and energy consumption. A robust framework is essential to perform these tasks, and in this thesis, we present a framework for collecting, processing, anonymizing, and predicting data.
Prior studies have employed different sensor data to capture various types of human activity. However, to capture and predict user behavioral patterns and intentions, we utilized commonly used sensors of smartphones, such as Bluetooth, Wi-Fi, and charging ports, which users frequently utilize. We developed an Android application to gather information from these wireless sensors and GPS. This research evaluated the accuracy and efficiency of three machine learning and four deep-learning architectures based on convolutional neural networks and LSTM variants. Additionally, we introduced a deep-learning architecture that employs ResNet and LSTM, which outperformed the traditional approach of combining convolutional networks with LSTM. The proposed model achieved high accuracy rates of 90.058% and 90.261% for predicting user locale and intent, respectively. Overall, this study highlights the potential of smartphone-based wireless sensor data and location services in predicting user activity and location. It emphasizes the importance of effectively utilizing deep learning techniques to process and analyze this data type and provides insight into leveraging smartphone data to predict human behavior in various application areas. | |
