Forecasting App

[NetZissou]

2021-06-01

The project kick-start meeting is on this afternoon. Some question to ask about:

• Users
  • Who will be mainly using this tool?
  • What kind of decision will the user make with this tool?
• Time Series Data
  • univariate or multivariate
  • minimum length of the series to be considered as valid observations
  • minimum length of the series to examine seasonality, trend, and make a prediction
  • ways of filling NAs, I prefer to handle this on the backend and provide the missing information
• Models
  • automodel in wicast
  • check collinearity of variables, normality of residuals, heteroscedasticity, outliers,

Potential functionalities include:

1. Data missing information
2. Time series components
3. Trend detection --> numeric
4. Seasonality detection --> potential reason for not detecting seanality
5. Predictions --> Explaining the model performance, prediction intervals

Note:

• Finalize most of the functionalities, and try out different user interfaces.
• Categorize instead of stacking up functionalities.
• It would be best to organize the functionalities step by step instead of putting them all in one basket
• Building Decision Trees
• Avoid unnecessary plots. Don't distract the user. Deliver your message most straightforwardly.

Name this tool and create a repo.