الفهرس | Only 14 pages are availabe for public view |
Abstract Nowadays, Smartphones are playing important role in human life as considered the primary communication tool. Additionally, the users of smartphones can perform variety tasks such as watching videos, playing games, listening to music, browsing the internet, etc. However, smartphone are battery based devices; therefore they have a limited amount of energy. The battery lifetime prediction can help the user optimizing the smartphone usage in such a way that can prolong the duration of the battery charge. In this thesis, we propose a data mining based system that is able to classify the users of mobile devices based on their usage patterns, into one of three classes, namely High, Medium, and Normal. Also, the proposed system can estimate the remaining battery lifetime of a mobile device. The proposed system includes two main phases: data preprocessing and data processing. In the data preprocessing phase, a set of operations are applied on the used dataset to make it ready for the next phase. These operations include parsing, handling missing data (by two methods: deleting missing values and compensating missing values by average imputation), normalization, statistical operations, and clustering using k-means clustering algorithm. In the data processing phase, both classification and prediction models are using a number of well-known data mining techniques. IX Abstract In the proposed system, four classification models have been used Naïve Bayes, multilayer perceptron, support vector machine, and J48 algorithms, respectively. The algorithms are applied on both datasets that are resulted from handling missing data methods. In addition, four prediction models have been used Naïve Bayes, multilayer perceptron, support vector regression, and linear regression. The algorithms are applied on both datasets resulted from handling missing data methods. |