Time Series

Non-continuous Data

RIMA, SARIMA, Exponential Smooth models might not work.   Gradient Boosting methods like CatBoost are good alternatives.  

Length

Nural network models such as LSTM need much more than 100 observations for training.  

Stationary Data

Non-Stationary Data

With Both Trend & Seasonality

Holt-Winters’ Exponential Smoothing

Trend Only (No Seasonality)

Seasonal Only (No Trend)