Getting Started with FormulaML

Welcome! This guide will help you get up and running with FormulaML in just a few minutes. By the end, you’ll have created your first machine learning model entirely within Excel.

Installation

Prerequisites

• Microsoft Excel 2016 or later (Windows or Mac)
• Active internet connection for initial setup
• Excel add-ins enabled in your Office settings

Installing FormulaML

1. Open Excel
2. Go to Insert → Get Add-ins (or Office Add-ins)
3. Search for “FormulaML” in the store
4. Click Add to install the add-in
5. Once installed, you’ll see FormulaML functions available with the ML. prefix

Verifying Installation

Test that FormulaML is working by typing this formula in any cell:

=ML.DATASETS.IRIS()

If you see <Dataset>, FormulaML is installed correctly!

Key Concepts

Before diving in, let’s understand three essential concepts:

1. Object Handles

FormulaML uses “object handles” to manage complex data structures in Excel. When you see something like <Dataset> or <LinearRegression>, that’s an object handle. Think of it as Excel’s way of storing a complete dataset or model in a single cell.

Example:

• Cell A1: =ML.DATASETS.IRIS() returns <Dataset>
• This handle represents the entire Iris dataset (150 rows, 5 columns)

2. The ML Workflow in Excel

The typical machine learning workflow in FormulaML follows these steps:

1. Load Data → Creates a DataFrame handle
2. Split Data → Separates features (X) and target (y)
3. Train Model → Uses ML.FIT to create a model handle
4. Make Predictions → Uses ML.PREDICT with new data
5. Evaluate → Check model performance with ML.EVAL.SCORE

3. Function Namespaces

All FormulaML functions are organized into logical namespaces:

• ML.DATASETS.* - Built-in datasets for learning
• ML.DATA.* - Data manipulation and exploration
• ML.REGRESSION.* - Regression algorithms
• ML.CLASSIFICATION.* - Classification algorithms
• ML.CLUSTERING.* - Clustering algorithms
• ML.EVAL.* - Model evaluation functions
• ML.PREPROCESSING.* - Data preparation tools

Your First Model

Let’s create a simple classification model using the famous Iris dataset. This dataset contains measurements of iris flowers and classifies them into three species.

Step 1: Load the Data

In cell A1, enter:

=ML.DATASETS.IRIS()

Result: <Dataset>

Step 2: Explore the Data

Let’s see what’s in our dataset. In cell A3, enter:

=ML.DATA.INFO(A1)

This shows the structure of your data: 150 samples with 4 features plus 1 target column.

To see a sample of actual data, in cell A5:

=ML.DATA.SAMPLE(A1, 5)

This displays 5 random rows from your dataset.

Step 3: Prepare Features and Target

We need to separate our features (flower measurements) from our target (species).

In cell B1 (select features - columns 0-3):

=ML.DATA.SELECT_COLUMNS(A1, {0,1,2,3})

In cell C1 (select target - column 4):

=ML.DATA.SELECT_COLUMNS(A1, 4)

Step 4: Split for Training and Testing

Good practice requires splitting data into training and testing sets.

In cell A10:

=ML.PREPROCESSING.TRAIN_TEST_SPLIT(B1, C1, 0.2, 42)

This splits 80% for training, 20% for testing. The result is an array of 4 handles:

• X_train (training features)
• X_test (test features)
• y_train (training target)
• y_test (test target)

Step 5: Create and Train a Model

Let’s use a Support Vector Machine (SVM) classifier. In cell A15:

=ML.CLASSIFICATION.SVM()

Result: <SVC> (untrained model)

Now train it with your data. In cell A16:

=ML.FIT(A15, INDEX(A10,1), INDEX(A10,3))

Result: <SVC> with 🧠 icon (trained model) This trains the SVM model using X_train and y_train from the split.

Step 6: Make Predictions

Use your trained model to predict on test data. In cell A18:

=ML.PREDICT(A16, INDEX(A10,2))

Result: Predictions for your test set

Step 7: Evaluate Performance

Check how well your model performs. In cell A20:

=ML.EVAL.SCORE(A16, INDEX(A10,2), INDEX(A10,4))

Result: A score between 0 and 1 (1 being perfect accuracy)

Congratulations! You’ve just built and evaluated your first machine learning model in Excel!

Understanding Your Results

What the Score Means

• Classification models: Score represents accuracy (0.95 = 95% correct predictions)
• Regression models: Score represents R² (how well the model fits the data)
• Scores closer to 1 indicate better performance

Object Handle Management

• Object handles persist during your Excel session
• Save your workbook to preserve handles between sessions
• Use descriptive cell labels to track what each handle represents

Next Steps

Now that you understand the basics, explore these areas:

📊 Work with Your Own Data

Learn to convert Excel data to DataFrames and apply ML to your datasets.

🤖 Try Different Algorithms

Explore other models like Linear Regression for numerical predictions or K-Means for clustering.

📈 Advanced Evaluation

Learn about cross-validation and hyperparameter tuning (Premium features).

💡 Complete Examples

Follow our step-by-step tutorials for real-world scenarios:

• Predicting Housing Prices
• Customer Segmentation
• Email Classification

Common Questions

“What if I get an error?”

Check our troubleshooting guide for common issues and solutions.

“Can I use my own data?”

Yes! Learn how in our Working with Data guide.

“Which algorithm should I use?”

See our Models and Algorithms guide for choosing the right model.

“How do I improve accuracy?”

Explore model evaluation techniques and advanced features.

Quick Reference Card

Ready to dive deeper? Continue to Working with Data →