# Contents

**Tools | Cross Validation | Structural Coefficient Analysis** computes a frequency value that measures how robust each arc is. This analysis is available in the **Validation Mode**, whenever a database is associated with the network.

Be careful when setting the lower boundary for the analysis, i.e. the **Minimum Structural Coefficient**. As you approach zero, networks can grow to become very complex and take a very long time to compute.

For analyzing the **Structural Coefficient**, the selected learning algorithm will be tested with different values of the **Structural Coefficient** within the interval you define. The network structure is then repeatedly learned using the associated database, using an increasing value for the **Structural Coefficient** with each iteration.

### Parameters

Simply select the learning algorithm you want to use, and set the minimum and maximum limits for the **Structural Coefficient** and the number of iterations to perform in the following dialog box:

The **Structural Coefficient** can range from 0 to 150.

Three measures can be computed at each iteration:

**Structure/Data Ratio**: use this in the context of**Unsupervised Learning**. It is the ratio between the default structural complexity (with a coefficient to 1) of the obtained networks and their data likelihood (with a coefficient to 0).**Target Precision in %**: use this for**Supervised Learning**, which requires a**Target Node**. The precision of the target prediction is computed for networks at each iteration.**Structure/Target's Precision**: use this for**Supervised Learning**when considering the trade-off between structural complexity and predictive precision.

The to-be-tested **Structural Coefficients** will range from the set minimum to the maximum, with an increment of (maximum - minimum) / number of iterations.

An output directory can be specified where all to-be-learned networks will be saved.

### Analysis Report

Once the networks are learned for each value of the **Structural Coefficient**, the following report is displayed:

This report is similar to the **Arc Confidence Report**, except for the last table. Here, there is an additional column indicating the **Maximum Structural Coefficient of the Structure**. This indicates, among all the coefficients tested, which value was the highest for which the structure was obtained.

The report can be saved in HTML format or printed. Furthermore, there are three options: **Network Comparison**, **Extract Network**, and **Curve**.

### Network Comparison

As the name implies, the **Network Comparison** button brings up a comparison of each learned network's graph structure. This allows for an easy visual interpretation of the report.

### Extract Network

Clicking **Extract Network** brings up a **Network Extraction Tool**. This allows you to pick a specific network structure as a function of the arc frequency thresholds you proved.

### Curve

Clicking the **Curve** button brings up a dialog box with options, which reflect what you set earlier when you started the analysis. You can chose to display one measure at a time.

**Structure/Data Ratio**: this plot helps you find a suitable value for the**Structural Coefficient**. Moving from right to left, chose the inflection point, just as the curve starts to rise sharply.

**Target's Precision in %:**

**Structure/Target Precision Ratio:**