Machine Learning Practical Workout | 5 Real-World Projects

Learn to use Google Colab without installing software, create a new notebook, run Python code, and manage files, datasets, and sharing options.

Explore the iris dataset through a train-test split (80/20) and build a k-nearest neighbor classifier to label testing data into three target variables.

Explore the iris dataset for flower species classification, using sepal length, sepal width, petal length, and petal width across 150 samples, with 50 samples per species: setosa, versicolor, and virginica.

Upload the iris dataset to Google Colab by creating a new notebook, connecting, and uploading the dataset from the folder, and enjoy GPU access from Google without buying hardware.

Import libraries in Google Colab: pandas as pd, numpy as np, and a visualization library like mitraclip; import k-nearest neighbors classifier, preprocessing, and train/test split for analysis.

Read the dataset with pandas and load it using read_csv with the provided path, then display the first five rows with iris.head() and the last five rows with iris.tail().

Explore the iris dataset with pandas by inspecting shape, columns, and values. Review the iris info, describe statistics, and note how to split data into x and y for modeling.

Split the dataset into features X (first four columns) and target Y (last column) for classification using iris data, preparing X and Y for model building and accuracy evaluation.

Learn how data normalization standardizes features using a standard scalar, applying fit_transform to training data and scaling test data, followed by an 80/20 train-test split.

Build a data model using a k-neighbors classifier with three neighbors, fit it on X_train and Y_train, then predict X_test and evaluate accuracy with Y_test.

Calculate the model's accuracy using the accuracy score and print the result, showing 96.67 percent accuracy with 97 correct predictions out of 100.

Classify flower species using the iris dataset in pandas, covering data loading, inspection, normalization, and a knn model. Achieve 97% accuracy with an 80/20 train-test split.

Explore a mushroom classification problem by building regression-based models with logistic regression and SVM, train on a subset of the dataset, and evaluate accuracy to compare models.

Explore a mushroom classification project using a 8124-row csv dataset with 23 features, identifying edible (E) vs poisonous (P) mushrooms through cap shape, color, gill attachment, and more.

Upload the mushroom dataset to Google Colab by clicking the folder icon, selecting upload, and loading the file; learn how to read from the dataset next.

Import essential libraries such as numpy, pandas, and scikit-learn preprocessing and models, including label encoder, standard scaler, logistic regression, and svm, and prepare the dataset with train_test_split.

Read a csv mushroom dataset by assigning data = read_csv('path'), copy and paste the file path, then preview the first five rows to reveal 23 columns.

Learn data preprocessing with label encoding using a label encoder to convert categorical columns into numeric labels and their mappings, then split into x and y for modeling.

Split the data into X (features) and Y (target 'class'), drop the class from X, then standardize X with StandardScaler to remove the mean and scale to unit variance.

Split the data into 80 percent training and 20 percent testing, using X, Y and X_train, with about 6499 training and 1625 testing rows.

Train logistic regression and support vector machine models, evaluate with test data, and compare scores to show SVM achieves higher accuracy (up to 100%) in classification projects.

Classify mushrooms as edible or poisonous by preprocessing a character-valued dataset with label encoding, then train and compare logistic regression and SVM models on an 80/20 split, achieving 100% accuracy.

Explore the problem of credit card fraud by building a logistic regression model on a credit card dataset, train, test, and evaluate accuracy to start the project.

Explore the credit card fraud dataset from Kaggle, with time, amount, and PCA-transformed features, highlighting an imbalanced binary target (0 no fraud, 1 fraud) and logistic regression for fraud detection.

Learn how to import a dataset into Google Colab by right-clicking a folder, selecting upload, and choosing the dataset path, as shown with a credit card dataset.

Import numpy as NP and pandas as dx to prepare the data, and import train_test_split, logistic regression, and accuracy_score from sklearn for model training and evaluation.

Learn to read the dataset by copying the file path, loading it into a data frame, and inspecting the first five rows and dataset information for null values.

Explore the distribution of normal and fraudulent transactions in the data, highlight the imbalance, and show how to split and shape normal and fraud subsets for model-ready data.

Balance the dataset by sampling an equal number of normal transactions to match the distribution, then concatenate with fraud data using pandas to yield 984 rows and 31 columns.

Split the dataset into X features and Y target by dropping the last column as the class, then prepare X and Y for training and testing in the next video.

Split data into train and test sets with 80% training and 20% testing, separating features x and labels y, with optional stratification and a fixed random state for reproducibility.

Train a model using logistic regression by fitting it to the training data X_train and y_train, then evaluate accuracy in the next video.

Assess the logistic regression model's accuracy score on the X test and compare predicted versus actual results in the card fraud prediction project. The model achieves 95.41 accuracy, indicating performance.

Explore a credit card fraud detection project that preprocesses data, balances an imbalanced dataset, and builds a logistic regression model achieving about 95.4% accuracy on a train-test split.

Apply a regression approach to a Big Mart sales dataset using an 80/20 train-test split, build an xgboost regression model, and evaluate the predictions with a score.

Explore the big mart sales dataset by examining features such as item identifier, weight, fat content, visibility, item type, MRP, and outlet attributes to understand item outlet sales.

learn to upload a dataset in Google Colab using the folder view, right-click upload, and select the big sales dataset; observe the progress indicator and note varying upload time.

Import essential libraries for the project, including numpy, pandas, and a visualization library, plus xgboost, along with sklearn preprocessing, train-test split, and metrics for encoding and regression.

Learn to read a dataset with pandas, convert to a dataframe, display the first five rows, inspect shape and info, and identify missing values using isnull for data preparation.

Learn how to handle missing values by replacing the item weight with its column mean, permanently updating the dataset, and setting up future handling of outlet size.

Learn how to impute missing outlet size values using the mode by outlet type, apply the transformation, and verify that there are no remaining missing values.

This lecture introduces the data analysis step, showing how to use describe to get mean item weight, standard deviation, counts, min, max, and percentiles from the dataset.

Visualize numerical and categorical features with seaborn, including distribution plots for item weight and MRP, sales, and count plots for outlet establishment.

Standardizes labeled categories in a dataset by mapping variants of low fat and regular to single labels, uses replace operations to unify values, and introduces label encoding for categorical data.

Apply label encoding to convert categorical features such as item identifier, item type, outlet identifier, and outlet features into numeric values using a fit_transform encoder, preparing features for model training.

Split the data into features x and target y, with item outlet sales as the target. Exclude the sales column from x and prepare for training and testing.

Split the data into training and testing sets with an 80/20 split using train_test_split, yielding X_train, X_test, y_train, and y_test, preparing for an XGBoost regression model in the next video.

Train a regression model using x_train and y_train, predict on both training and test data, and evaluate with r-squared scores to assess model performance.

Load a 12x12 sales dataset in Google Colab, fill missing values with mean and mode, encode categorical features, visualize distributions, and evaluate a regression model with R-squared around 58.7.

Explore heart disease prediction by training a logistic regression model on a dataset, evaluate accuracy on the test set, and predict labels for new data as normal or diseased.

Explore the heart disease prediction dataset: a 14-column csv with 13 features such as age, sex, chest pain, blood pressure, cholesterol, fasting glucose, max heart rate, plus the target variable.

Import numpy, pandas, logistic regression, train_test_split, and accuracy_score to prepare the project for data reading and analysis.

Read and load the heart disease dataset, then preview head and tail. Check info, missing values, describe statistics, and analyze the target distribution.

Split the dataset into features X and target Y by dropping the target column, then go on to split into training and testing data.

Split data into training and testing. Use 80% for training and 20% for testing, with X and Y to form X_train, X_test, y_train, and y_test, random state 2, and stratification.

Train a logistic regression model on x_train and y_train, then evaluate its accuracy using the accuracy score, reporting an 81.96% accuracy.

Learn to build a predictive model by preparing new input data, reshaping it for the model, and using model.predict to decide if a person has heart disease.

The project analyzes a heart disease dataset with 303 rows and 14 columns, where 1 indicates disease and 0 normal, using logistic regression on an 80/20 split with 81.96 accuracy.

Explore the UCI machine learning repository to access datasets for ML and data science, browse by category (classification, regression, clustering, others) and data types (categorical, numerical, mixed), across research areas.

Explore Kaggle datasets and sign in to access items like Netflix, Alcohol, FIFA, and African Study, then use notebooks and challenges to practice regression and data science.