Predicting the Sale Price of Bulldozers

Objectives:

• To predict the future sale price of bulldozers, given its characteristics and previous samples
• To build and train a model using a random forest regressor with time-series data
• To tune hyperparameters using RandomizedSearchCV

Steps:

1. Data pre-processing
• Import data and parsing dates
• Sort dataframe by saledate
• Add datetime parameters for saledate column
• Convert string to categories
• Fill missing values
• Turn categorical variables into numbers
• Split data into train/validation sets
2. Building a model
• Build a machine learning model using random forest regressor
3. Training a model
• Train a model using trained dataset
4. Model evaluation
• Evaluate a model with root mean squared log error (RMSLE) metric
5. Tuning hyperparameters
• Tune hyerparameters with RandomizedSearchCV
• Train a model with the best hyperparamters
6. Deployment
• Make predictions on test data
• Extract Feature Importance

Outcomes:

Heart Disease Classification

Objectives:

• To classify the presence of heart disease in patients based on various health metrics using a classification model
• To build a model using machine learning techniques, including hyperparameter tuning and model evaluation
• To compare different classification models and select the best one based on performance metrics for heart disease prediction

Steps:

1. Data pre-processing
• Load data
• Data exploration (Exploratory Data Analysis: EDA)
2. Modelling
• Model comparison
• Hyperparameter tuning
3. Hyperparameter tuning with RandomizedSearchCV
4. Hyperparameter tuning with GridSearchCV
5. Model evaluation
• Calculate evaluation metrics using cross-validation
• Feature importance
6. Experimentation
• Evaluation metric

Outcomes:

Multi-class Dog Breed Classification

Objectives:

• To identify the dog breed given a dog image using a multi-class image classifier
• To build a model using deep learning and transfer learning with TensorFlow and TensorFlow Hub
• To train and evaluate the model with prediction probabilities of 120 dog breeds

Steps:

1. Data pre-processing
• Get data ready: images and labels
• Create our own validation set
• Preprocess images: turning into Tensors
• Turn our data into batches
2. Building a model
• Build a model using Keras sequential model and MobileNetV2
3. Training a model
• Train a model using trained dataset
4. Evaluating a model
• Make predictions on validation data using a trained model
5. Training a model with full data
6. Deployment
• Make predictions on test dataset and custom images

Outcomes: