Mileage prediction is calculated using a machine learning method with an additive model and the MCMC (Markov Chain Monte Carlo) algorithm. The model is trained considering trends, seasonality factors, and a series of regressive components.
You can view the mileage forecast on the chart in Reports. The chart displays a maximum of 12 months forecast, but the app calculates a 5-year forecast for each car. Below is an example of mileage from January 1st of the current year and the forecast, as well as a blue line of mileage for the previous year for this car (the “Comparison” feature is available with the “Premium” extension). Thus, you can compare the mileage for the past (2022) and current (2023) years, including the forecast.
When you add expense, income, fuel/charge and reminder entries with a date in the future, the upcoming section will appear on the “Timeline” tab. If at the same time the mileage field is empty, the app will calculate the predicted mileage and show it on the “Timeline” tab.
If you create a reminder that should be triggered by a certain mileage in the future (without a date specified), the app will calculate the forecast date.
All forecast values (mileage and date) are displayed in gray. Forecasting is calculated only for future periods. If you have not entered the mileage in the records of expenses, income, refueling / charges for a date that has already passed, then the app will calculate the approximate mileage using the average values method.
Read more about the forecast feature in our blog.