Building Model for Crime Pattern Analysis Through Machine Learning Using Predictive Analytics

Abstract

Crime has a big impact in both the human lives and the society’s growth, which needs to be addressed and controlled. Machine learning algorithms as the fanciest technology to assist decision makers in policy making has proven its reliability in showing unseen patterns in crime. This research aims to examine the capability of trees and ensemble trees in classifying crime through model development. Experiments were done to enhance the capability of the ensembles in both classification and regression. Feature extraction like synthetic minority oversampling technique was applied in order to address the problem in the imbalanced data. Different metrics relevant to classification and regression were considered in evaluating the performance of each model used. With the use of different metrics, Gradient boosted tree was found to have better classification capability in crime dataset after outperforming decision tree and random forest in both classification and regression problem. Furthermore, random forest was also found to have a promising capability in classification by regression. Therefore, it is highly recommended that this ensemble algorithm be further examined and considered in developing model in other datasets.

Prediction of Senior High School Students’ Performance in a State University: An Educational Data Mining Approach

“…This ensemble algorithm is composed of numerous structured-trees * ( ) + where * + is the independent but identical vector distributed to cast vote in searching common class for each x . Furthermore, in the experiment conducted by [16] random forest shows a promising result in classification by regression problems which leads to be an interesting area to study for prediction.…”