Research Article

Quantitative Analysis and Prediction of Global Terrorist Attacks Based on Machine Learning

Table 4

Comparison of 5 algorithms for predicting terrorist groups with different attack frequencies.

Summary of data recordsNumber of terrorist attacks (range)≥1000≥500≥100≥50≥5
Number of terrorist organizations1932122210936
Total number of terrorist attacks5020058520781078433994871

AccuracyDecision trees0.9826690.9586470.8783770.8546360.796164
Bagging0.9607570.9319890.8333120.7998580.740620
Random forests0.9830680.9682160.9044940.8811950.835687
ExtraTrees0.9792830.9595010.8866980.8603270.803225
XGBoost0.9834660.9716340.8539240.7914390.698567

PrecisionDecision trees0.9769500.9282420.7875210.7456480.478754
Bagging0.9459560.9321520.7712530.6856090.347113
Random forests0.9792320.9577270.8475590.8173840.520747
ExtraTrees0.9733270.9421590.8112420.7612730.476816
XGBoost0.9784060.9572460.7522350.5234900.126893

RecallDecision trees0.9760730.9295540.7860340.7379200.512161
Bagging0.9400900.8581060.6051440.5233390.304769
Random forests0.9747430.9341400.7852650.7396190.511993
ExtraTrees0.9700250.9262340.7615100.7085500.469377
XGBoost0.9760590.9449040.7465250.5378580.138689

F1 scoreDecision trees0.9764970.9286030.7841850.7364700.481310
Bagging0.9410570.8751510.6330590.5511910.305712
Random forests0.9765870.9428830.8054880.7545970.502975
ExtraTrees0.9716090.9327980.7790960.7238340.459432
XGBoost0.9771180.9500110.7459250.5238000.130349