AI experiments advanced options

<< Click to Display Table of Contents >>

Navigation:  Cloud applications > Bizagi Artificial Intelligence > Working with AI experiments >

AI experiments advanced options

Overview

Once you have created an AI experiment and generated the model, and when evaluating the resulting accuracy of the prediction, you may switch to advanced options in order to modify the algorithm employed or further explore other combinations.

Note that these options will most likely require a good degree of expertise on the machine learning subject as a whole.

For information about generating the model to interpret results, refer to Working with AI experiments.

 

Advanced options

Switch to the advanced options by clicking on Advanced View:

 

Cloud_AI_switchdvanced

 

While in the advanced view, you are presented with a graphical view of how Bizagi Artificial Intelligence has determined the best treatment and algorithm for the predictive service.

Such graphical view presents the steps carried out in order (each step represented by a box), and order from top to bottom.

 

Regarding this configuration and graphical view, there are 2 different flows you may want to reconfigure:

 

1. The training flow.

The data treatment, algorithm and training of the model, as employed by Bizagi during the model generation step.

Toggling the training flow (on/off) is done by clicking the icon having 2 arrows:

 

Cloud_AI_expert_ds

 

2. The predictive flow.

The series of steps employed by Bizagi at runtime, whenever the predictive analysis service is invoked.

Toggling the predictive flow (on/off) is done by clicking the icon having 2 arrows:

 

Cloud_AI_predictiveFlow

 

Interpreting the flows

You may follow up the sequence of each flow and review the specific configuration for each of the steps by clicking on its box.

To the right, you will see the configuration Bizagi Artificial Intelligence has applied, as determined best for the use case.

 

The image below shows as an example of a training flow and how is the first step configured.

Such step is of the type DataSource (typically the first one of a training flow), which represents that step where data is obtained.

Notice it is configured with indications of the target Dataset and the attributes to use (selected features):

 

Cloud_AI_interpretds

 

By clicking Feature selection for this type of step, note you may redefine the selected features, as described at Working with AI experiments.

 

For other steps, such as Clean Data, note that different settings apply.

The image below shows as an example, that the upper Clean Data step applies for the pregnant, plasma, bpressure, skinthickness, 2h_insulin, BMI, and age attributes and that its clean policy (for empty values) considers replacing them by the average value of the set.

 

Cloud_AI_interpretclean

 

You may notice there may be more than one Clean Data step and this is because each would have a different treatment for empty values for the selected attributes.

 

Recall that for replacing empty values, you also have the possibilities presented according to the data type of the feature, as shown at Working with AI experiments.

 

Cloud_AI_expert_trainingflow2

 

For other steps, you may note that multiple sequence lines can generate from them, or to them.

In the case of Split Data, note that a left fraction indicates what fraction of data goes to one step (and hence, the remaining fraction ---that which completes up until 1.0-- goes to another step).

The image below shows how Split Data sends 70% percent of its data to the Train Model step (to be able to achieve a best certainty for a prediction, a preselected model is reinforced), and the remaining 30% goes to Validate Model (to verify and tests results).

 

Cloud_AI_interpretsplit

 

Notice you have also a step indicating which algorithm is selected for the model (the image above shows Multiple Linear Regression).

In a similar way and with usually fewer number of steps, a predictive flow is interpreted as well.

 

Modifying the flows

In order to modify how any flow is defined, rely on the boxes found at the left panel, as grouped per their function.

To include any of these, simply drag and drop them into the flow:

 

Cloud_AI_expert_trainingflow

 

To delete a step to connect other ones you add to the flow, select a step and use the keyboard's delete key.

Then rely on the endpoints that each shape has so that you can drag and drop a sequence line from the source shape (its endpoint):

 

Cloud_AI_expert_trainingflow5

 

And into the target shape's endpoint:

 

Cloud_AI_expert_trainingflow4

 

In the end, you should have redefined a flow which is consistent in the sense that it connects adequate all steps you include in it:

 

Cloud_AI_expert_trainingflow3

 

Notice you may use Model evaluations such as: Score model, Validate model, or Cross validation:

 

Cloud_AI_expert_modelev

 

Notice you may use Data Transformations such as: Word count, Random split, SMOTE, Select Columns, Remove Columns, Normalize Numbers, Make categorical, Join Data sets, or Clean missing values (Clean Data):

 

Cloud_AI_expert_datatransf

 

Notice you may use Machine learning algorithms such as: C45, ID3, Linear SVM Classifier, Multiple Linear Regression, Logistic Regression or Train Model:

 

Cloud_AI_expert_mlalgorithms

 

Finally and when you are done, generate the model and ensure you review or test thoroughly its results, so that you can publish that experiment if satisfactory, as described at Publishing an AI experiment.