ADS Capstone Chronicles Revised

21

vu as el idd at ot i oi dnewn ti itfhyi nt heea oc hp ttirma ianl ihnygpdeartpaasreat mweatse r s faoprper aocahc hm, tohdeenl wt hiet hRi nM eSaEc wh ams ocdael cl iunl ga t e d o n tshheo wt rsa itnhien gr easnudl t itne gs todpattiamsiezte. dT a b l e 5 . 1 b e l o w hmyopdeerlpi nagr aamp pe treorasc ahne ds . R M S E s f o r b o t h

5 Results and Findings 5.1 Model Performance: RMSE At rlal imn eodd et wl s i mc ee, no tni coen eo dn itnh eT at rbal ien4i n. 1g wd aetraes e t go ennc ee roant etdh ef rtorma i nMi nogd edlai nt ags Ae tpgpernoearcaht e1da fnrdo m Modeling Approach 2. Five-fold cross

Table 5.1 Results from Grid Search through Hyperparameters using Five-Fold Cross-Validation. For each Modeling Approach’s Training Dataset, the Optimal Hyperparameters and RMSE Values are Shown. Model Approach 1 Optimal Hyperpara meters Approach 1 Training RMSE Approach 1 Test RMSE Approach 2 Optimal Hyperpara meters Approach 2 Training RMSE Approach 2 Test RMSE Linear Regression N/A 0.0001238 0.0001238 N/A 0.0001237 0.0001238 Ridge alpha: 10.0 0.0001238 0.0001238 alpha: 10.0 0.0001237 0.0001238 Lasso alpha: 0.1 0.0001778 0.0001782 alpha: 0.1 0.0001779 0.0001783 ElasticNet al 1l p_ rhaat:i o0:. 10,. 1 0.0001778 0.0001782 alpha: 0.1, l1_ratio: 0.1 0.0001779 0.0001783 XGB Regressor l 0 e .1 ar , ning_rate: m20a,x_depth: n20_estimators: 0.0000480 0.0000484 learning_rate: 0m.1a,x_depth: 2 n 0 _e , stimators: 20 0.0000478 0.0000491 LightGBM Regressor l 0 e .1 ar , ning_rate: m20a,x_depth: n20_estimators: 0.0000311 0.0000359 learning_rate: 0m.1a,x_depth: 2 n 0 _e , stimators: 20 0.0000300 0.0000355 Support Vector Ce p: s0i.l1o, n : 1 0.0011862 0.0011862 C: 0.1, epsilon: 1 0.0011862 0.0011863

112