devi2k20 / crime-predictor-chicago Goto Github PK

This project forked from shanbhag10/crime-predictor-chicago

One of the most important questions people ask before moving to a new place or traveling is: Is this locality safe? We answer this question using co-location pattern mining. Our crime predictor gives the possibility of all crimes that can happen given the presence of certain geographical entities and certain crimes in the locality. Given a set of boolean spatial features, the co-location pattern discovery process finds the subsets of features frequently located together. We have used a transaction free approach for this using Haversine distance as a metric for finding proximity neighborhood and measuring all spatial autocorrelations. We also propose an improvement to distance measure approach while calculating R-proximity neighborhood using boundary window. We used distance and participation thresholds for determining the correlation between features and crimes. Our algorithm exploits the anti-monotone property and optimizes the co-location pattern discovery algorithm by prevalence based pruning.

Home Page: https://drive.google.com/open?id=1CFsYQRAEYG0MNZv7T6wbJnsA_6lREAIq