Trend Analysis of Terrorist Incidents in Pakistan-1996-2017

Abstract

Pakistan is one of the highly effected countries of the world from terrorism, this project is consisting of three different parts which are Trend Analysis, Terrorist Groups Classification and Incident Date Prediction. The Trend analysis part is a graphical representation of what Pakistan has gone through in the last four decades. The data used in this project is the subset if GTD (Global Terrorism Database) related to Pakistan. It contains the terrorist incidents that happened in Pakistan from 1970 to 2017. Trend analysis shows the trends such as numbers of incidents that happened per year, affected provinces, type of weapons used, type of attacks and many more like most active terrorist groups, most numbers of target types etc. Jupyter notebook is used for the algorithm pandas and matplotlib is used for the plotting of the data after some preprocessing. The Terrorist Groups Classification analyzed all the available data of terrorist activities in Pakistan, tried several algorithms and techniques of machine learning, used the GTD data related to Pakistan for the prediction of the responsible groups (perpetrator) of a given terrorist incident, using data of the attack like type of the attack, targets and weapons used also the city and province, date like year and month and more available features of the attack. It is found out that the Support vector machine (SVM) algorithm with sigmoid kernel predicts groups with higher than 85% accuracy. The Incident Date Prediction is the sophisticated algorithm which used machine learning techniques like regressions and RNNs (Recurrent Neural Network). The data used in this project has three separate columns for year, month and day for date. so, combining these columns in a single number used for regression and RNN has done via a simple algorithm. The single numbered data is then fed into Autoregression and RNN which accurately predict the next incident date with not more than 4 days tolerance. This project can help in reporting terrorism dark ages of the country as well as can augment the investigating process and help the agencies with an extra factor to consider, which can help to improve the overall security of Pakistan as it is transitioning into a digital Pakistan.