import pandas as pd

import numpy as np

from sklearn.cluster import DBSCAN

from sklearn.neighbors import KernelDensity

from geopy.distance import geodesic

import matplotlib.pyplot as plt

# Load the data

crime_data = pd.read_csv(‘REDACTED’)

crime_data[‘Date’] = pd.to_datetime(crime_data[‘Date’], errors=’coerce’)

# Filter for the last 3 years

recent_crime_data = crime_data[crime_data[‘Date’].dt.year >= 2017]

# Kernel Density Estimation for identifying the probable location

location_data = recent_crime_data[[‘Latitude’, ‘Longitude’]].dropna()

X = location_data.values

kde = KernelDensity(bandwidth=0.01, metric=’haversine’, kernel=’gaussian’, algorithm=’ball_tree’)

kde.fit(np.radians(X))

latitude_range = (location_data[‘Latitude’].min(), location_data[‘Latitude’].max())

longitude_range = (location_data[‘Longitude’].min(), location_data[‘Longitude’].max())

x_mesh, y_mesh = np.linspace(latitude_range[0], latitude_range[1], 100), np.linspace(longitude_range[0], longitude_range[1], 100)

X_mesh, Y_mesh = np.meshgrid(x_mesh, y_mesh)

XY_samples = np.vstack([X_mesh.ravel(), Y_mesh.ravel()]).T

Z = np.exp(kde.score_samples(np.radians(XY_samples)))

predicted_idx = np.unravel_index(np.argmax(Z, axis=None), Z.shape)

predicted_lat_lon = (x_mesh[predicted_idx[0]], y_mesh[predicted_idx[1]])

# Predict the most likely crime type

radius = 2.0

filtered_recent_crimes = recent_crime_data.dropna(subset=[‘Latitude’, ‘Longitude’])

nearby_crimes_extended = filtered_recent_crimes[filtered_recent_crimes.apply(lambda row: geodesic((row[‘Latitude’], row[‘Longitude’]), predicted_lat_lon).kilometers <= radius, axis=1)]

crime_counts_extended = nearby_crimes_extended[‘Primary Type’].value_counts()

most_likely_crime_extended = crime_counts_extended.idxmax() if not crime_counts_extended.empty else filtered_recent_crimes[‘Primary Type’].value_counts().idxmax()

# Predict the exact spot for the theft

theft_nearby = nearby_crimes_extended[nearby_crimes_extended[‘Primary Type’] == ‘THEFT’]

most_recent_theft = theft_nearby[theft_nearby[‘Date’] == theft_nearby[‘Date’].max()]

exact_spot = most_recent_theft[[‘Latitude’, ‘Longitude’]].iloc[0] if not most_recent_theft.empty else None

print(f”Predicted Location: {predicted_lat_lon}”)

print(f”Most Likely Crime Type: {most_likely_crime_extended}”)

print(f”Exact Spot for Theft: {exact_spot}”)