Analyzing Data for Machine Learning

Before training a model, XGENIA can analyze your dataset to detect the problem type, suggest the best target column, identify quality issues, and recommend preprocessing steps.

What you will learn in this guide

• How to use the AI chat to analyze data
• What insights the analysis provides
• How to use the Auto ML Analyzer pro node

Using the AI Chat

The simplest way to analyze data is through the AI assistant. Ask it to analyze your data:

Analyze this customer data for churn prediction:
[paste your data or describe it]

The AI calls hf_analyze_data under the hood, which sends your data to the ML Coordinator for analysis.

Analysis Output

The tool returns structured insights:

Field	Description
problemType	Classification, regression, or clustering
suggestedTarget	The column most likely to be the prediction target
suggestedFeatures	Columns that are good predictors
hfTaskType	The corresponding HuggingFace task (e.g. tabular-classification)
qualityIssues	Missing values, outliers, class imbalance warnings
preprocessingSteps	Recommended data preparation steps
columns	Per-column statistics (type, uniqueness, missing rate)

Using the Auto ML Analyzer Pro Node

For visual workflows, drag the Auto ML Analyzer node onto your canvas.

Inputs

Port	Type	Description
data	Array/Object	Your dataset (array of records)
mlServerUrl	String	ML server URL (default: http://localhost:3001)
context	String	What you want to predict (e.g. "customer churn")
hfToken	String	HuggingFace API token
Analyze	Signal	Trigger the analysis

Outputs

Port	Type	Description
insights	Object	Full analysis results
problemType	String	Detected problem type
targetColumn	String	Suggested target column
suggestedFeatures	Array	Recommended feature columns
dataQualityIssues	Array	Data quality warnings
preprocessingSteps	Array	Recommended preprocessing
hfTaskType	String	HuggingFace task type
error	String	Error message if analysis fails

Example Workflow

1. Connect a Query Records node or REST node to provide data
2. Wire the data output to the Analyzer's data input
3. Set the context to describe your prediction goal
4. Trigger the Analyze signal
5. Use the outputs to configure training (target column, features, task type)

Retention Analysis

For customer retention specifically, use the Client Retention Analyzer node. It provides specialized outputs:

• Churn rate — percentage of churned customers
• Risk factors — columns correlated with churn, ranked by significance
• Recommendations — actionable suggestions to reduce churn

Wire it to a Retention Action Engine node to automatically generate action plans based on the risk factors.