Hands-on Machine Learning with Python & ChatGPT

Explore how machine learning, a subset of artificial intelligence, uses supervised and unsupervised data to learn patterns and predict outcomes with models like linear regression and K-means, plus feature engineering.

Clean data starts the workflow by addressing missing values, outliers, and errors. Prepare features through data manipulation, scaling, and encoding; develop, evaluate, and deploy models for regression, classification, or clustering.

Learn to load an Excel dataset in a Jupyter notebook, install and import pandas, and use read_excel to display the first five rows.

Identify and impute missing values in a dataframe using pandas and scikit-learn, employing isnull and simple imputer with median for numeric data and the most frequent value for categorical data.

Identify and fix inconsistent data types in a dataset by verifying numeric costs as float. Use pandas to inspect dtypes, locate non-numeric values, and remove them before conversion.

Remove inconsistent values from a numeric column using Pandas, then cast columns to the correct types (float for cost, date time 64 for dates) to ensure clean data.

Identify and remove duplicated data using a Python data frame workflow, with code generated by ChatGPT, producing a clean data frame free of duplicates.

Learn to manipulate data by sorting a dataset by a numeric value using python code, pandas and sort_values, with ascending or descending options, and see insights through sorted costs.

Filter datasets with conditions in Python to extract data for France or for France and the games category, and remove France using not equal operator, all demonstrated via ChatGPT help.

Learn how to add an extra variable to an existing dataset by merging on a key (order id) with pandas merge, and filter to matching ids for the refund value.

Learn to concatenate supplementary data vertically with pandas pd.concat, merging masked data with extra rows from extra_data.xls to expand the final dataframe.

learn to apply feature engineering to create new variables from customer data, including median sales, median cost, and purchase frequency, using a Python pivot table.

Extract day, month, and year from a date variable in a dataframe by converting to datetime and using dt attributes, then create and inspect new day, month, and year columns.

Learn to encode a categorical variable by assigning numeric values, giving loyal customers 1 and regular customers 0, via a new variable customer type encoded using Pandas replace.

Create dummy variables for nominal data using pandas get_dummies to convert country, category, and device type into 0/1 indicators for machine learning models.

Scale the data with a standard scaler to balance feature magnitudes and improve model performance, using refund as the target and other variables as features.

Split the scaled features and the refund target into train and test sets to measure model accuracy. Use a 20% test size with a fixed random state for consistency.

Build a linear regression model to predict refund amounts using train and test data, evaluate with mean absolute percentage error, and visualize predictions against actual values and residuals.

Hands-on with decision tree regression, this lesson demonstrates building and evaluating a Python model to predict refund amounts, showing improvements over linear regression using MAPE and visual residual plots.

Develop and evaluate a random forest regressor to predict the refund amount, comparing its mean absolute percentage error to the linear regression and decision tree models.

Explore building a support vector regression model to predict refund amounts, compare it with linear regression, decision tree, and random forest using mean absolute percentage error, and observe SVR performance.

Explore building an XGBoost regression model to predict refund amounts and compare it with linear regression, decision trees, random forests, and SVR using mean absolute percentage error.

Predict customer type using a logistic regression classifier on encoded features, perform data preprocessing and train-test split, and evaluate with a confusion matrix and accuracy.

Explore the decision tree classification model to predict customer type (regular vs loyal), compare its accuracy and confusion matrix with logistic regression, and learn practical Python code steps for implementation.

Deploy a random forest classification model to predict customer types, train on data, test with a confusion matrix and heatmap, achieving 59% accuracy and 41% error.

Explore k nearest neighbors classification to predict customer type (regular vs loyal), compare with random forest, test multiple k values, and visualize accuracy and confusion matrices.

apply Lightgbm to classify customer type as regular or loyal in a binary model, train on data, predict on test, and assess with confusion matrix, accuracy, and error rate.

Apply k-means clustering to uncover inner characteristics in customer data and segment into regular and loyal customers using median sales, median cost, and purchase frequency.

Learn to diagnose and debug Python errors with ChatGPT, impute missing education values, reshape data to 2D arrays, and create advanced age-by-education visualizations.