Machine Learning No-Code Approach: Using Azure ML Studio

Build your first machine learning model for Titanic survival classification using Azure ML Studio, covering data upload, feature selection, train-test split, training, and evaluation with a guided demonstration.

Split the data into training and test with the split data module, using 70% training and 30% testing, and save your work with a fixed random seed.

Train a two-class decision forest model by selecting the algorithm, supplying training data from the split, and predicting survived with a column selector, then visualize and score on new data.

Improve your classification model by analyzing data, selecting features, cleaning misinformation, and comparing algorithms, then deploy your model via web service and Excel in a no-code Azure ML Studio workflow.

Revisit your first model and inspect data quality using mean, median, standard deviation, and missing values. Use the Summarised Data module to examine quartiles and feature usefulness for predictions.

Experiment with a satcom machine learning algorithm and compare two-class models in Azure ML Studio, evaluating accuracy and area under the curve to select and deploy the best model.

Deploy the refined model by setting up and validating a predictive web service. Connect an Excel file to submit data and retrieve scored predictions from the web service.

Refine iris flower experiment by stratifying data by class and re-running to raise accuracy from 88.89% to 97.78%, comparing multiclass logistic regression with another model and inspecting the confusion matrix.

Deploy your first machine learning model without code, using Titanic data to apply data summary, feature selection, cleaning, stratification, and hyperparameter tuning for classification; then explore regression for house prices.

Create a Milbourne project, upload datasets, and conduct exploratory data analysis to understand features, select columns, and address missing values for modeling.

Convert string features to categorical using the edit metadata tool in manipulation, turning type, coastal area, and region name into categorical features after cleaning missing data.

Explore mean absolute error and relative absolute error calculations using actual price and predicted price, with a practical excel workflow to compute these indicators from two columns in a dataset.

Explore an Azure ML Studio no-code workflow solving the Iowa housing market exercise by selecting key features and predicting sale price, with mean absolute error 23,758 and 40% relative error.

Reassess selected features, tune hyperparameters, and compare algorithms to improve a regression model while noting variance. Deploy the model live via Exo and test with data using a no-excel approach.

Use future based feature selection in Azure ML Studio to assess feature-to-price correlations, prune weak features, keep strong ones, and compare model results across experiments.

Compare regression algorithms by testing a boosted decision three regressor against the prior model, evaluating mean absolute error on test data, and preparing deployment in Excel.

Explore additional algorithms, test new models, and compare results; sample datasets and gallery experiments help you build and run your own projects.