# Explainable AI charts

The explainable AI charts in Vizpads offer users a detailed view of how machine learning models derive their predictions. This bridges the gap between black-box AI predictions and tangible, understandable factors that influence each outcome.

{% hint style="danger" %}
Please note that explainable AI charts are exclusively available in Vizpads and are not part of the Search module.
{% endhint %}

### **Components of Explainable AI charts**

1. This chart presents a visual representation of the influencing factors for a machine learning model's prediction.

* **Prediction**: Shows the predicted outcome of the machine learning model. It gives a snapshot of what will occur based on the data and factors.
* **Accuracy**: Displays the accuracy percentage of the model's prediction. It helps gauge the reliability of the AI's insights
* **Factors**: Lists down the factors affecting the prediction, ranked by their importance.
* **Values:** Each factor comes with a value. A positive value indicates the factor positively impacts the prediction, while a negative value means it hampers it.

{% hint style="info" %}
The column used for prediction will be displayed at the top (next to **Prediction** and **Accuracy).**
{% endhint %}

<figure><img src="https://content.gitbook.com/content/s16h5onryWtbaHwBa10b/blobs/P0GFe8MosaR04cfCuxYT/image.png" alt="" width="375"><figcaption><p>Explainable AI chart</p></figcaption></figure>

2. To access a detailed breakdown of the prediction, click on any row within the chart.
3. Upon selection, the left panel will update to display a list of contributing factors along with their respective values, offering a clear insight into the prediction's reasons.

By understanding the influencing factors, users can get a clear insight into the 'why' behind every AI prediction and make informed decisions.

### **How to create a new explainable AI chart?**

1. Navigate to **Explore -> Create new Vizpad** or open an existing Vizpad.
2. Click on **Add Chart -> Explainable AI**.
3. In the **Configuration** pane on the right, choose the required **ML Model**. This model determines the predictions and factors shown in the Explainable AI Chart.

<figure><img src="https://files.helpdocs.io/6dnnwn52e3/articles/x2fjkbyqfh/1698678388719/image.png" alt="" width="375"><figcaption><p>Configuration of explainable AI chart</p></figcaption></figure>

1. From that model, select the required **Business View** from a list of Business Views used by the ML-Model.
2. The target columns selected for prediction and accuracy will be auto-populated under **Columns**, and highlighted in yellow with a lock sign displayed adjacent to each column name.
3. **Sort by**: Double-click the columns by which the chart needs to be sorted or drag and drop the required columns under the **Sort by** field.
4. **Unused:** The columns that cannot be used for the chart but are selected will be listed here.
5. Enable **Pagination** if you want data to be displayed on separate pages. Provide the required number to be shown per page in the **Show Rows** field.

### **Factors Sorting**

The "**Factors Sorting**" section allows users to customize the order in which these factors are displayed, based on their values. By toggling between these options, users can gain a clearer understanding of which factors play a dominant role in AI's predictions and how they relate to each other in magnitude and direction.

<figure><img src="https://files.helpdocs.io/6dnnwn52e3/articles/x2fjkbyqfh/1698680196853/image.png" alt="" width="375"><figcaption><p>Factors sorting section</p></figcaption></figure>

#### **1. Value Type:**

* **Absolute**: When you select this option, the factors will be sorted based on their absolute values. It ignores whether the influence of a factor is positive or negative. This mode focuses solely on the magnitude or strength of the influence, not its direction. For example, if you have factors with values of -10, 5, and -8, they'll be sorted as 10, 8, and 5 in absolute sorting.
* **Real**: In this mode, factors are sorted by their genuine values, including their positive or negative influence. This mode gives you a holistic view by considering both the magnitude and the direction of the influence. Continuing with the previous example, they would be sorted as 5, -8, and -10 in real sorting.

#### **2. Sort Direction:**

The "Sort Direction" determines the order (ascending or descending) in which the factors are presented.

* **Decreasing**: When selected, the factors will be displayed from the highest to the lowest value, based on the chosen value type (either **Absolute** or **Real**).
* **Increasing**: Opposite to the decreasing option, factors are listed from the smallest (or most negative) to the largest (or most positive) based on the selected value type.