@@ -190,3 +190,33 @@ Missing data
190190
191191Using mean imputing;
192192-------------------------------------------------------------------------------------------------------------------------------
193+ 5. Tesing the Model's Accuracy
194+ -------------------------------
195+ * Statistics are only data. We define what is good or bad;
196+ * Performance Improvement Options
197+ -------------------------------
198+ a. Adjust current algorithm;
199+ b. Get more data or improve data;
200+ c. Improve training;
201+ d. Switch algorithms;
202+ 203+ * Random Forest
204+ ----------------
205+ -> Ensemble Algorithm;
206+ -> Fits multiple trees with subsets of data;
207+ -> Average tree results to improve performance and control overfitting;
208+ 209+ - Train with training data: y = x1 + w2 * (x2)^3 + w3 * (x3)^8
210+ - complex decision boundary;
211+ - good fit of training data;
212+ - poor fit of test data;
213+ - Overfitting;
214+ 215+ Fixing Overfitting
216+ ------------------
217+ * Regularization hyperparameter
218+ y = x1 + w2 * (x2)^3 + w3 * (x3)^8 - f(W)/(lambda)
219+ y=x1+w2 * x2^3+w3 * x3^8−f(W)/λ
220+ - Cross validation;
221+ - Bias - variance trade-off;
222+ - Sacrifice some perfection for better overall performance;
0 commit comments