Optimal Approach for Data Pre-Processing of Time Series Data
Narayana Darapaneni, David Pratap, B G Sudha, M S Ullas, Anwesh Reddy Paduri · 2024
The paper proposes an optimal approach for data extraction from the “Rythu Bazar” official website Department of Agricultural Marketing, Govt. of Telangana (tsmarketing.in). The scope of data to work for this project was the “Hyderabad” market only for 26 commodities (perishables) starting from the year 2017 to December 2019. An attempt has been made to apply the algorithm for the other commodities like potato, beetroot, and tomato. The Results of this algorithm were compared to the results from backfilling EWMA approaches. As a novelty to the existing approach, a normalized RMSE algorithm was built along with the MAPE as an additional metric. As per the business understanding of data, the batch size was set to a week instead of using the number provided by the original authors. Different approaches for scaling the data were tested on the dataset and the results were very convincing.