python - Predict number of rows using Machine Learnnig -
i have bank data of around 4 years of different branches. trying predict number of rows in daily , hourly level. have issue_datetime (year, month, day, hour) important features. applied different regression techniques (linear, decision trees, random forest, xgb) using graph lab not better accuracy. thinking set threshold based on past data taking mean of counts in daily, monthly level after removing outliers , set threshold. best approach?
since have 1d time series data, should relatively easy graph data , interesting patterns.
once establish there non-stationary aspects data, class of models wanting check out first auto-regressive models, possibly seasonal additions. arima models pretty standard time-series data. http://www.seanabu.com/2016/03/22/time-series-seasonal-arima-model-in-python/
Comments
Post a Comment