This is the implementaion of methods proposed in the research mentioned below: https://www.researchgate.net/publication/336059621_Investigating_Auxiliary_Information_in_Multivariate_Time_Series_Forecasting_Models
Forecasting multivariate time-series always been a hot domain either from the era of statistical models and now in the time of deep learning models. Our research includes analyzing best performing models and mainly focused on finding the role of auxiliary information in forecasting models. Our work Investigates the integration of supplemental information for corresponding multivariate time-series models for more than one labels. This extra inform- ation which we call as auxiliary input can be fed as vital features in respect- ive model architecture and lead to a performance gain. Although this study has well-known use cases like in parking, stock market, energy consumption and many more, our research methodology is applied to predict parking oc- cupancy at multiple locations by several experiment and techniques using different models in this work. Google has the first form of high scale solution for the same use case with a combination of crowd-sourcing and Machine Learning models with resourceful data. Our work is mainly covering Ma- chine Learning domain to accomplish a similar task, so it can be seen as a subset solution to compare with the Google parking difficulty prediction. Performance gain analysis by integration of the auxiliary features in models design is being done with other data-sets as well than parking availability to support our research.