The problem She-Farer solves

Consider a user who is visiting an unknown city and wants to visit the Hotspots nearby his location. The usual choice is to google 'Places near me' , ask friends or post questions on social platforms like Quora, which mostly accounts to biased suggestions and therefore does'nt always guarantee a satisfactory trip. "SHE-Farer" is a Recommendation Engine provided in the form of a web application, that can answer the same questions but considers the opinions of diversified population of a wider demography. We have about 10,000 user's data synthetically generated based on a normal distribution, and have clustered them based on euclidean distance of measure of their features. We clustered the user data into 10 different clusters, where 10 (no. of clusters) was obtained from the Elbow curve method of plotting. So, when a new user enters the system, we map him to one of the existing clusters, which contains users who have similar preference to that of the new user. Based on the historic data of ratings given to different places by the users in this cluster, we predict what our current user will rate those places based on the ML algorithm, Collaborative Filtering. The places with the highest predicted ratings is recommended to the user and plotted as a route in map.

Challenges we ran into

The first challenge was aquiring the huge user data that is required to build this recommender engine.This was made easy by the Here API which could fetch details of places for the specified locations. The second challenge was handling and manageing a huge dataset of 10,000 users and about 300 places. The Surprise library in python came handy to excecute the collaborative filtering algorithm on this huge dataset. The only issue that took for ever to be solved was plotting the shortest route between the places recommended and preferred by the user.