.

This work examines solutions to the growing problem of spam and fraudulent messages that are prevalent in the mobile phone industry today. It begins with an examination of some common methods for detecting spam messages such as: the Rule-based method and the Statistical Learning method using Naive Bayes approach. This work specifically explores Naive-Bayes classifier for categorizing messages based on their resemblance with words that feature in other spam and non-spam messages in the training set, thereby reducing the number of spams that get through to the end user and completely eliminate false positives (messages that are misclassified as spam). Incorporated in the dataset for this project is the SMS Spam Corpus v.0.1 Big. It has 1,002 SMS ham (legitimate) messages and 322 spam messages. For the initial training and testing of the Naive-Bayes classifier, Python 2.7 interpreter, Sublime Text text-editor, Plotly for data visualization and comparing results were used while Java Development Kit, Android SDK, Android Studio and Android Emulator were used for deployment. It can be concluded that using a spam threshold of 0.7 along with adjustments to the Naive Bayes algorithm, we obtained some desirable results. In an attempt to improve on the method used in this work, we are currently working on how to use hybridized machine learning algorithms for detecting Spam messages in mobile devices.
Keywords: SMS; Spam filtering; Naive Bayesian filter; threshold
INTRODUCTION
Spam