The problem Bus Chacha solves

Smart City is the hottest topic of discussion currently and without a doubt , public transportation forms the backbone of a city. While Delhi Metro is an amazing concept for public transportation , it doesn't ensure door-to-door public transporation due to infrastructural limits. To combat this , Delhi metro feeder Bus service came into play but what was a failure as it used to operate at random locations. What we have done is that based upon dataset which gives us geographical locations of people who use delhi metro the most , we employed a machine learning algorithm to find the most suitable bus stops so as to cater to these target audience. It is a step forward to door-to-door public transporation and is of extreme practicality. It will definitely minimize the traffic and chaos around metro stations as people living a little away from metro station have to arrange an auto/ola/uber for themselves to reach metro stations which inturn contributes to otherwise huge traffic in the city. We propose running of buses at frequent intervals from metro station to these ideal bus stops suggested by our machine learning algorithm so that these people who use Delhi metro on a very frequent basis actually use it and chaos around metro stations could be done away with it. It is also a government aid since Delhi metro feeder bus service as a concept came into existence 10 years back but didnt bcome successful but with this algorithm since we target the rigt audience , we are sure to reap out maximum benefits. For the users , we have developed an android application which detects their current location , asks them for their destination location and suggests them the nearest bus stops while they commute to and from a metro station to their homes/offices.

Challenges we ran into

The Machine Learning algorithm starts with a random set of centres which gets optimised through thorough training . The bug that arose was the algorithm started to depend upon how the initial points were selected . If the initial points selected were completely vague , the suggested bus stops didn't seem ideal . Hence we use python library scikit learn which studies the dataset and ensures it starts with the best of points which are in suitable range and predicted stations are accurately identified.