| Popis: |
Jammu & Kashmir, despite being one of India’s most pristine regions, is also known for the threats it faces due to frequent terrorist attacks. Forecasting of potential terrorist activity becomes an important task that can help increase safety. We collect data of terrorist activity in Jammu & Kashmir from the South Asian Terrorism Portal and process it to form datasets consisting of batches of three different sizes. We then employ one machine learning model (Random Forest) and four deep learning models (Multilayer Perceptron, Convolutional Neural Network, Long Short-Term Memory and Bidirectional Long Short-Term Memory) to forecast terrorist activity frequency using regression, and we compare the results on the basis of Mean Absolute Error and Coefficient of Determination. From our analysis we establish that terrorist activity frequency can be forecasted using machine and deep learning methodologies. We observe that recurrent neural network models perform the best, and models generally perform better when given more historic features. |
