Development Stages
Stage 1
Status: Completed
Objective: Collect race data and transform it to a usable structure. Visualize the race data and present in a manner that users can understand. Create multiple ways to view the data in order to draw conclusion from a chart.
Notes: The transformation of the data requires turning the time from a string to float. For the data visualization, fix the y-axis to better allow for comparisons to be made. Created a zoomed in graph to allow for lap time comparisons to be made.
Stage 2
Status: Completed
Objective: Add comparison charts for teammates in the data visualization to compare the performance of drivers in the same car. Develop the framework for a multiple linear regression model and run it.
Notes: The comparison charts are useful in comparing drivers and we will continue to do this. We plan to expand the comparison to rivals. The biggest change to the comparison was ensuring the different drivers were clearly differentiated in the plot. For the multiple linear regression model, the initial model was to simple and did not properly work. The initial results can be seen on the Bahrain post-race report.
Stage 3
Status: In-Progress
Objective: Implement a machine learning model into the multiple linear regression model to boost accuracy and make more precise predictions. Allow for users to submit their predictions and get predicted lap times to compare for the race.
Notes: The implementation of the machine learning algorithm greatly increased the effectiveness of the prediction model and reduced the variability when compared to just the multiple linear regression model. The user submission has been created and is currently live at the Prediction page.
Stage 4
Status: Future Work
Objective: Improve the prediction model through data engineering. Implement new features in the prediction model to see how it impacts the lap times and work on making the data more aligned with the race. Develop a page for the model so users can upload their strategy and instantly get the results on their strategy.
Notes: Experiment with the existing features such as weight and make it non-linear when a safety car occurs. Experiment with implementing DRS activations in a lap to the data or a proximity to car feature which factors in the effects of dirty and/or close racing to the lap time.