Inspiration
When playing Apex Legends, there was a noticeable trend in where the zone was going. Combined with the recent backlash against pro-player data mining zones to gain the upper hand, we thought it would be beneficial to make a Machine Learning AI model to represent this trend.
What it does
This project starts with a website that allows users to traverse three different maps of the current season. Users can click on a given point on the map that represents their Zone 1 center point. Then it would make a request to the flask python backend to generate the resulting zones based on the point and KNN ML algorithm to display in a result page.
How we built it
We first cleaned the datasets because the code we used to parse the information didn't support the clear ring image. Then manually input the zone locations into a database to use in the training of the KNN model. Using the KNN where K is equal to the square root of the number of data entries divided by 2. Then for the training model, it trains sequentially for 100 x 100 matrices for the maps and their corresponding zones based on the initial zone location. The front end is built with ReactJS and the back end is with Python on Flask.
Challenges we ran into
Attempting to parse the information and cleaning the datasets itself took over 13 hours to complete and make it really short on time. It didn't really help that after cleaning the dataset, the center point was still off for many of the datasets, making us spend more time correcting the values. Furthermore, the algorithm training and modeling was a long and tedious process that had many errors and bugs, but eventually, that was resolved. Otherwise, the project isn't really completed because there is a compatibility issue with Flask and ReactJS where the request for generating subsequent zones would not go through when it is because React and Flask just cannot handle that. By the time this was discovered, there was no more time for workarounds or alternatives.
Accomplishments that we're proud of
None of us had experience with what we were doing in terms of back end and front end, but we still made a lot of good progress. We also managed to write out the ML KNN algorithm and completed the training code.
What we learned
A lot of Python and ReactJS
What's next for Apex Legends Zone Predictor
The next thing is to fix the connecting end points between the front end and the back end. Then continue to extend the functionality to consider other variable just as squads alive, heat maps, etc.
Log in or sign up for Devpost to join the conversation.