Embedded Files
A new data set was received from the Irrigation Department after contacting Eng. Aruna Samarathunga.
This data set contain water level of Kalu ganga taken from a sensor at rathnapura town area from January 2020 to January 2021 at a sample interval of 1 hour. Sample of the data set is given below.
Using this dataset univariate analysis was conducted. Autocorrelation plot for this data set is given below.
Magnified ACF plots are given below
By looking at this plot, lags up to seven hours can be taken to the model without a problem because they have a good correlation. (greater than 0.9)
Auto Regressive (AR) , Moving Average (MA) and ARMA models are created considering only one lag and predicted the water level.
Models performance as follows
Performance of AR Model
Performance of AR Model
RMSE = 0.00472
Performance of MA Model
Performance of MA Model
RMSE = 0.54167
Performance of ARMA Model
Performance of ARMA Model
RMSE = 0.02933
Downside with all above models is forecasting can done only for 1 hour ahead. If we need, further forecast, previous forecasted values has to be used as inputs for the models.
Page updated
Google Sites
Report abuse