- Created a machine learning model that classifies whether breast cancer is Malignant or Benign.
- Analyzed dataset with 569 observations and 33 columns
- Processed the data for making it suitable of better machine learning
- Optimized Logistic, and Random Forest Regressors using RandomSearchCV and cross validation to reach the best model.
- Made a sample prediction for a given observation
Python Version: 3.10.7
Packages: pandas, numpy, sklearn, matplotlib, seaborn
Dataset contains 32 features that are computed from a digitized image of a fine needle aspirate (FNA) of a breast mass. They describe characteristics of the cell nuclei present in the image.
Link to dataset - https://www.kaggle.com/datasets/uciml/breast-cancer-wisconsin-data
Used count plot for finding out whether the given dataset is balanced. I looked at the distributions of the features and analyzed correlation of the variables for finding out multicollinearity. Below are a few highlights from exploratory data analysis.
Count Plot of Output
Correlation Heatmap
After collecting the data, I needed to clean it up so that it was usable for our model. I made the following changes to the variables:
- Removed unnecessary columns
- Encoded categorical variables
- Removed features with multicollinearity
- Extracted features and target
- Standardized the dataset
First I split the data into train and tests sets with a test size of 20%.
I tried two different models and evaluated them using Accuracy score. Fine tuned model with best accuracy for better machine learning process.
Machine Learning models:
- Logistic Regression – chose this model because our problem is binary classification
- Random Forest – with default parameters
The Logistic regression model outperformed the Random forest model on the test and validation sets.
- Random Forest : Accuracy = 94.73 %
- Linear Regression: Accuracy = 92.98 %
Chose Logistic regression as final model this problem and fine tuned its hyperparameters.
- Tuned Logistic regression : Accuracy = 94.73 %, Precision = 91.48 %, Recall = 95.55 %, F1 Score = 93.47 %
Final model gave same accuracy as the model with default parameters, while precision score was little lower the recall and f1 score improved. The model made 108 correct predictions and 6 wrong predictions on the test data.
I made a sample prediction with the Tuned Logistic regression model with the given features:
radius_mean = 11.13, texture_mean = 16.62, smoothness_mean = 0.08151, compactness_mean = 0.03834, symmetry_mean = 0.1511, fractal_dimension_mean = 0.06148, radius_se = 0.1415, texture_se = 0.9671, smoothness_se = 0.005883, compactness_se = 0.006263, concave points_se = 0.006189, symmetry_se = 0.02009, symmetry_worst = 0.2383
The machine learning model predicted that the cancer with given observations is Benign