Global Terrorism Data Analysis

Predictng Terrorist "Group" given the data features.

Road map

1- Feature engineering :

a- Adding attributes to make more sensible attribute b- Removing the outliers i.e terrorists group which has occurrence less than 10 in the dataset

2- EDA :

a-Histogram Analysis : Observing the frequency distribution of numerical attributes. b-Scatter plot Analysis : Geographical distribution of casualities c- Observing the most active terrorist Organizations : This shows that the dataset is highly imbalanced d- Wordcloud : On Motives,Target,Types,Summary

3- Data Preparation :

a- Imputing nan values of categorical data points by their mode(most frequent) b- Applying pipeline to numerical features(imputer,scaling) and categorical features(1-hot encoding) c- Splitting training and test features and labels

4- Handling Imbalanced dataset:

a- Using SMOTE(Synthetic Minority Over-sampling Technique): This creates the oversampling of minority classes considering the neighbours of the class b- Using class_weight (During training) : This makes the model more bias towards the minority by penalizing heavily for the error happening in the case of minority class then to the majority class

5- Using Random Forest classifier

6- F1 Score : 0.719373911056525(avg:micro);0.6913907382270787(avg:weighted)