hungpham2002 / non-linear-regression-to-multivariable-non-linear-regression

Tabular Data Project: Sales Prediction. Derived from the idea of Linear regression extend to non-linear regression and Mutilvariate Non-linear regression

Non-Linear-Regression-to-Multivariable-non-Linear-Regression

• Preprocessing data (dealing with null, ...)
• EDA (Explore Data Analysis): Correlation between features and labels, Linear function or non-linear function, 2-degree or 3-degree,...
• Representation: Data transformation by using Label Encoding / One hot encoding, ...
• Modeling: training and evaluation
• Deployment

Important Note:

• Degree Choice: if too simple, model would not flexible enough (underfit), if too complicate, model would overfit
• 
• Good method for choice of the degree: k-fold cross-validation
  • Example: suppose we have 1000 samples: 800 training and 200 testing
    • focus on training set, testing set only be used to evaluate model.
    • The original dataset is divided into K equal-sized subsets or folds.
    • The model is trained and evaluated K times.
    • In each iteration, K-1 folds are used for training the model, and the remaining fold is used for testing/validation.
    • 
    • The performance metric (such as accuracy or mean squared error) is computed for each iteration.
    • The final performance measure is obtained by averaging the results from all K iterations.
    • Choose the degree which has the lowest out of sample error
• Disadvantages: