Data Science

Crime Against Women in India: A Data-Driven Approach to Understanding and Combating Gender-Based Violence

Introduction

Crimes against women in India represent a deeply entrenched societal issue, rooted in centuries of cultural and historical practices. Despite significant legal reforms post-independence, these crimes remain pervasive, driven by entrenched patriarchal norms and systemic challenges.

For instance, the dowry system—originally a form of inheritance—has transformed into a symbol of exploitation, resulting in various forms of abuse, including dowry-related deaths. Addressing such systemic issues requires a nuanced understanding of the historical, cultural, and legal factors perpetuating these crimes.

The Cultural Context

India's diverse regional norms and societal expectations shape gender roles, often reinforcing harmful stereotypes. This cultural backdrop influences reporting patterns and judicial outcomes. For instance, ingrained stigmas deter victims from reporting crimes, while systemic inefficiencies lead to low conviction rates. The interplay between these factors underscores the need for a comprehensive, data-driven approach to combating gender-based violence.

Problem Statement

The increasing rate of crimes against women necessitates a deeper analysis of crime patterns and judicial outcomes. This project focuses on: 1.⁠ ⁠Trend Analysis: Examining state-wise trends and gaps in judicial proceedings. 2.⁠ ⁠Prediction: Forecasting future crime rates and conviction probabilities. 3.⁠ ⁠Evaluation: Highlighting discrepancies between reported and convicted cases to address judicial inefficiencies.

Objective

By leveraging data-driven techniques, this project aims to: •⁠ ⁠Analyze crime trends and identify correlations between factors like state and type of crime. •⁠ ⁠Predict the number of reported cases and the likelihood of convictions. •⁠ ⁠Provide actionable insights for policymakers, law enforcement, and NGOs.

Data Overview

The project utilized datasets from Kaggle: •⁠ ⁠Cases_under_crime_against_women.csv: Detailing crimes reported from 2001 to 2010. •⁠ ⁠Arrests_under_crime_against_women.csv: Providing arrest details over the same period.

A merged dataset with 34 columns and 4165 rows was created for analysis. Key preprocessing steps included: •⁠ ⁠Handling discrepancies in ⁠ Group_Name ⁠ and ⁠ Sub_Group_Name ⁠ columns. •⁠ ⁠Standardizing labels for consistent analysis.

Exploratory Data Analysis (EDA)

EDA revealed significant trends, such as:

Variations in crime types across states.
Discrepancies in reported versus convicted cases.

Model Development and Predictions

The project utilized machine learning models for two key tasks:

1. Cases Reported Prediction

⁠Random Forest and SARIMAX models were employed for trend analysis and forecasting. Performance insights:

Random Forest captured complex patterns effectively, outperforming other models.

2. Cases Convicted Prediction •⁠ ⁠Models included ARIMA, K-Nearest Neighbors (KNN), Lasso Regression, Multiple Linear Regression (MLR), and Random Forest. •⁠ ⁠Performance Comparison:

ARIMA: Struggled due to limited data, resulting in overfitting.
KNN: Moderate accuracy but sensitive to feature variations.
Lasso Regression: Reduced overfitting and improved interpretability.
Random Forest: Achieved the lowest RMSE (168.0), demonstrating robustness in handling nonlinear relationships.

Research Implications

The findings have far-reaching implications: 1.⁠ ⁠Social Impact: Highlighting systemic challenges can drive societal change. 2.⁠ ⁠Judicial Analysis: Identifying inefficiencies aids in reforming judicial processes. 3.⁠ ⁠Policy Development: Insights can guide resource allocation and proactive policymaking.

Call to Action

Addressing crimes against women requires collaborative efforts: •⁠ ⁠Policy Makers: Use data-driven insights to craft targeted interventions. •⁠ ⁠Law Enforcement: Improve conviction rates by addressing procedural delays. •⁠ ⁠NGOs and Activists: Raise awareness and support victims through evidence-based advocacy. •⁠ ⁠Technology Developers: Innovate tools for real-time reporting and data monitoring.

Conclusion

This project underscores the potential of data-driven methodologies in tackling gender-based violence. By bridging gaps in understanding and action, we can pave the way for a safer, more equitable society for women in India.