Nowadays, Spam or known as unsolicited in E-mails has become a main problem and it is difficult to discover by self. Spam showed raise by internet users rapidly day by day. However, machine learning algorithms can ability to classification and detection unsolicited emails or non-spam to avoid harm or any risk during communication by email via organizations, companies, or personalities. Therefore, it is most important to detect fraud emails. In this paper we focused to give background about emails and spam. Next we compared machine learning algorithms in our work and we have obtained the Naïve Bayes algorithms are the better in accuracy and precision. Thus, we have made preprocessing that such as cleaning data, Data transformation, Data Integration, and Data Reduction for instance Stop words, Tokenization, and Stemming. However, we have applied the Naïve Bayes to the classification of either spam or non-spam.