Service User Manual for Data Visualization

Introduction

This service allows you to generate an interactive HTML report for dataset exploration using the ydata-profiling library. The report can be generated as:

Standard exploratory report,
Time series report (temporal EDA),
Stratified report by target class (binary classes only).

The service is useful for performing quick exploratory data analysis, visualizing variable distributions, missing values, correlations, and other descriptive statistics.

Service Features

1. Exploratory Report

If no optional parameters are provided, an interactive report with a generic exploratory analysis of the dataset is generated.

2. Time Series Report

If the timestamp_column is specified, the service uses the temporal mode of ydata-profiling (tsmode=True), ordering the dataset rows according to the indicated column. This type of report is suitable for temporal data.

3. Stratified Report by Target

If the target_column is specified, the service generates a report for each distinct value of the target column, comparing the different profiles generated. If the classes in the specified target column are more than two, the service will fail and will not produce the expected result.

4. Customization via Direct Parameters

Through the direct_args_to_ProfileReport argument, you can specify additional parameters in JSON format that are directly compatible with the ProfileReport function, to customize the service behavior (e.g., disable some report sections, change the title, etc.).

Service Usage

Dataset Upload

The user must upload a compatible tabular dataset (e.g., in CSV format).

Parameter Configuration

timestamp_column (optional): name of the column to use for temporal ordering.
target_column (optional): name of the target column for stratification.
direct_args_to_ProfileReport (optional): JSON dictionary of parameters for the ProfileReport function.

If both timestamp_column and target_column are provided, the temporal mode takes precedence.

Output

The service produces an HTML file titled profiling_report.html, viewable within the platform through the application media available upon completion of the service run.