Complex models with denoising

      Linear    ElasticNet  SVR     xgboost
==========================================================================
raw
---------------------------------------------------------------------------
MSE   1197.8    1167.3       736.0   840.3
R2    0.6818    0.6899      0.8045  0.7768
score inf       inf         inf     inf

ma/fe
---------------------------------------------------------------------------
MSE   1057.7     959.1       966.6   610.3
R2    0.6950    0.7234      0.7212  0.8240
score inf       inf         inf     inf

The fit results of the complex model have also been improved by denoising sensor readings with moving average except for SVR. Actually, for SVR and xgboost, hyper-parameter tuning to be performed to get optimal model. The results above have been obtained with the default parameter settings for SVR and xgboost models.