Detection of Cybercrimes in Online Connections by the Data Mining Approach

At recent years, online social network sites have been popular dramatically. Cybercrimes use from social media as a new platform at acceptation of some types of computer crimes like phishing, spamming, malware spread and cyber harassment. In this research, we will improve the function of detecting cybercrime with the help of useful information in the messages. Choosing the best features with high separation. Strength between cyber harassment tweets and none cyber harassment is a complex activity which extremely needs substantially effort in making Machine Learning Model. In this way, we compare function of five classification methods Naive Bayes, Support Vector Machine, Decision Tree, k-Nearest Neighbor and Neural Network under five different tuning in order to selecting the best adjustment for suggested features. Also, we have improved C and Sigma parameters by using the bat, genetics and particle swarm algorithms. Additionally, we have compared five classification methods with default parameters and parameters obtained with optimization algorithms. Finally, we have shown that bat algorithm has had the best performance among other optimization algorithms. According to the research we did the most accuracy with the SVM model to 86.56 and the highest precision to 87.14.