Theft behaviors analysis in Chicago

Overview

The project is conducted to explore the character traits of theft from a significant amount of raw data - one of the complicated issues in data analysis. To address this challenge, I implement an automated data processing that cleans and normalizes the datasets, then reports to the DBA. Particularly, the DBA uploads the dataset on the server and receives a notification of the status of these pieces of data without performing any actions. Furthermore, in this project, I present a series of approaches to explore data insight. First, I pursue OLAP queries to understand the overview of the theft’s behaviors. Second, I implement a linear regression model to determine what factors affect the increase in theft behaviors in Chicago city. Finally, I build machine learning to predict behavior that is classified as theft or not. You can read the full report here.

Architecture

The project is structured by Cross-industry standard process for data mining (CRISP-DM):

1. Business understanding
2. Data understanding
3. Data preparation
4. Modeling
5. Evaluation

The data warehouse architecture

Dashboard

Some dashboards are built using Power BI.

The dashboard examines the number of theft crimes in Chicago

The dashboard shows underlying trend in criminal behavior

The dashboard shows crime ware over areas

Data mining

Using linear regression

Coefficients of variables using linear and logistic regression

Positive and negative coefficients of variables

Relationship between the predictor variables and response variable

Data modeling

Build model with different predictor variables