Project Story: Crimelens India
Crimelens India is a data-driven exploration of district-level crime data aimed at uncovering patterns, disparities, and predictive insights across the country. Our goal was to empower policymakers, law enforcement, and citizens with an intuitive tool to visualize crime trends, identify risk zones, and plan proactive strategies based on data.
Inspiration
We were inspired by the rising concerns around public safety, especially the lack of accessible and granular data analysis tools in the public domain. News headlines often highlight crime statistics without context — we wanted to dive deeper to answer:
- Where and when do different crimes peak?
- What types of crimes are most common in specific regions?
- Can we forecast and classify high-risk districts?
These questions led us to build Crimelens India — a platform combining exploratory data analysis, visual storytelling, and machine learning.
What We Learned
During the project, we learned:
- The importance of understanding domain knowledge around crime categorization and IPC sections.
- How crime rates are influenced by demographics, geography, and reporting practices.
- The effectiveness of data storytelling, dashboards, and geospatial visuals in making complex findings accessible.
- The role of model interpretability and feature selection in achieving reliable predictive results.
How We Built It
- Data Cleaning & Preparation: We cleaned and merged crime records, handled missing values, and added population-based proxies.
- Data Integration: Integrated data from various sources (crime stats, census data, shapefiles) and resolved inconsistencies in district names and formats.
- Exploratory Data Analysis: Identified high-crime districts, state-level variations, category-wise breakdowns, and seasonal crime spikes.
-Machine Learning & Statistical Modeling:
- K-Means Clustering: Grouped districts based on similar crime profiles.
- Random Forest Classifier: Classified districts as high-crime vs. low-crime.
- Linear Regression: Used for predicting future crime rates based on historical trends.
- Hyperparameter Tuning: Experimented with different parameters to optimize performance.
Challenges Faced
- Data Quality Issues: Missing entries, inconsistent district names, and varying formats across years.
- Data Integration: Merging datasets across different formats required careful mapping and transformation.
- Domain Understanding: Legal jargon and IPC sections required time to decode and classify appropriately.
- Urban vs. Rural Analysis: Due to missing urban-rural flags, we used population size as a proxy.
- Model Performance: Addressed class imbalance and overfitting during classification tasks.
Final Thought
Crimelens India demonstrates how data science can illuminate complex social issues like crime. With smarter tools and transparent insights, we can enable timely interventions and build safer, more informed communities. This project is a step toward using open data and machine learning to drive positive societal impact.
Log in or sign up for Devpost to join the conversation.