kanish-h-h / monthly_armed_robbery

Time Series Analysis of Monthly Robberies in Boston (1966-1974)

Project Overview

This project focuses on analyzing the monthly robbery incidents in Boston from 1966 to 1974 using a manually configured ARIMA model. The aim is to identify patterns and make future predictions based on historical data. The project model is saved at models/.

Dataset

The dataset consists of monthly records of robbery incidents in Boston between 1966 and 1974. The data is sourced from and is stored in the file data/Robberies.csv.

Project Structure

The project is organized into the following directories and files:

.pickle_files/
    
data/ 
    nb_data/
        
    Robberies.csv
        
    armed_robberies.zip
        
images/
    
model/
    
notebook/
	
README.md

Installation

Clone the repository:

git clone https://github.com/kanish-h-h/monthly_armed_robbery.git
cd monthly_armed_robbery

Pickle file

model.pkl can be found at folder .pickle_files/ for storing the model as pkl.

Usage

1. Data Preparation: Start by running the Armed_Robberies_in_Boston.ipynb notebook to clean and prepare the dataset and converts those into datasets.csv, validation.csv and stationary.csv.
2. ARIMA Modeling: Use the Armed_Robberies_in_Boston.ipynb notebook to configure and train the ARIMA model on the prepared data over manually configured.
3. Fine Tuning: Fine tuning is done via Grid Search method for finding the optimal p, d and q, for the ARIMA.

Results

The results of the analysis, including time series plots and ARIMA model diagnostics, are saved in the images/ directory.

Result on validation dataset

Time Series Plot

Residual Error

Conclusion

The ARIMA model provides a robust framework for analyzing and predicting time series data. This project demonstrates how to manually configure and apply an ARIMA model to real-world data, yielding insights into the patterns of robbery incidents in Boston over the observed period.