Project Overview:
In this part of the exercise, the objective is to implement regularized logistic regression to predict whether microchips from a fabrication plant pass quality assurance (QA). During QA, each microchip undergoes various tests to ensure it is functioning correctly. As the product manager of the factory, the goal is to determine whether microchips should be accepted or rejected based on the results of two different tests. A dataset containing test results on past microchips is available to build a logistic regression model to aid in decision-making.
Contributions:
-
📅 Situation: The task is to build a predictive model to classify microchips as pass or fail based on QA test results. The need arises from the QA process in the fabrication plant and the desire to automate decision-making regarding microchip acceptance.
-
🎯 Task: Implement regularized logistic regression to classify microchips as accepted or rejected based on QA test results. This involves utilizing a dataset of past microchip test results to train the logistic regression model.
-
⚙️ Action:
- Visualizing the Data: Explore and visualize the dataset to understand the distribution and characteristics of the QA test results.
- Feature Mapping: Enhance the model's ability to fit the data by creating additional features from each data point. Features are mapped into all polynomial terms of the QA test scores up to the sixth power, resulting in a 28-dimensional feature vector.
- Logistic Regression Implementation: Code the logistic regression algorithm from scratch to perform binary classification of microchips as accepted or rejected.
- Regularization: Incorporate regularization to prevent overfitting and improve the generalization of the logistic regression model.
-
📈 Result:
- The logistic regression classifier achieved an accuracy of 83% in classifying microchips as pass or fail based on QA test results.
- By using regularization, the model was able to mitigate overfitting and improve its performance on unseen data.
The implementation of regularized logistic regression provides a robust solution for automating the decision-making process regarding microchip acceptance in the fabrication plant's QA process.