Open Source Project - PyEmits

this is an development agenda for ideaInsight. it is inspired by Darts, which provides user-friendly implementation on timeseries data, but it doesn't support high resolution data forecast like seconds base. Seeing the date by date works on timeseries data, i would like to design a library for easy manipulation on timeseries data

The features shall include:

1. metrics

2. data processing

   1. selecting data in array

   2. reformat data into ratio, for better prediction. startpoint data = 1

   3. data cleansing

      1. pyod?

   4. splitting

3. back-testing

   1. simulating historical forecast, validate the data

4. forecast module

   1. lstm

   2. ml

   3. stats module

5. cross-validation

   1. sklearn cross validation

6. evaluate module/plotting lib

   1. provide confidence interval

7. data analytics

   1. descriptive data analytics

   2. diagnostic data analytics

   3. predictive data analytics

   4. commandative data analytics

8. continuous evaluator

9. window aggregation

10. dashboarding

11. blending/stacking module to advance the timeseries forecasting

also inspired by:

1. greykit

1. changepoint detection

2. model summary

3. seaonality

2. pytorch-forecasting

3. darts

4. pyaf

5. orbit

6. kats by facebook