0

I am building my first R project about forecasting rental car sales. I have 2 data sets which have different calculations and consist of daily and monthly data for 10 years.

Monthly dataframe enter image description here

Daily dataframe enter image description here

I want to forecast daily, monthly and quarter-year. I have used time series and auto.arima modeling for observing the trend and seasonality of the datasets. In addition to arima modeling, ı used some feature values such as holidays, weather situations as regressors and gbm (gradient boosting model). but, the results of the prediction is not good enough. enter image description here

enter image description here

green line: auto.arima forecasting with regressors

blue line: auto.arima forecasting without regressors

red line:. real data

What is the best model to enhance results?

Community
  • 1
Emre Aktürk
  • 101
  • 2
  • 3
    Possible duplicate of [How to know that your machine learning problem is hopeless?](https://stats.stackexchange.com/questions/222179/how-to-know-that-your-machine-learning-problem-is-hopeless) – Stephan Kolassa Feb 05 '18 at 08:34

1 Answers1

1

Welcome to the site!

There is this package named forecastHybrid, this might help you in getting better results. There are couple of reasons for suggesting this package:

  1. It ensemble of models like auto.arima,ets , thetam, nnetar etc.
  2. It gives equal weight-age to all the models, we can even changes the weights.
  3. Even if 1 model doesn't perform well, you can customize and take only the necessary models.
  4. Finally, you can use the external factors and take ensemble of them.

For implementation you can go through this link, they explained it with example.

Let me know if you need any additional information.

Toros91
  • 113
  • 6
  • It would be better if you could (a) give more details about how you think this would help the OP, (b) expand your answer with details about just what that package does as if the links go dead your answer will become valueless. – mdewey Feb 05 '18 at 10:11