Model Management

[ 1 Overview ] [ 2 What permissions do I need? ] [ 3 What is Machine Learning? ] [ 4 Upload a model in Model Management ] [ 5 Working with models ]

Overview

Model management allows you to manage your models through Machine Learning. Via Model Management you will be able to create, read, update and delete models. This will help you enhance the security of your data, and predict any outcomes using historical data.

Learn more about Machine Learning

Want to check some Machine Learning tutorials and learn how to upload and use some interesting models to your Devo domain? Check this GitHub repository and these docs to learn more. Get in touch with us if you have any questions.

What permissions do I need?

In order to work with Model Management you need the Manage level of the Models permission. Having only the View level will allow you to see them but not create them or edit existing ones (more info about permission here).

I can’t see these permissions

Note that if these permissions are not available in your domain by default, you can contact Devo to enable them.

What is Machine Learning?

Machine Learning is a type of artificial intelligence that allows Devo to become more accurate at predicting outcomes without being explicitly programmed. It uses historical data as input to predict new output values. There are three basic approaches:

Supervised Learning is a collection of Machine Learning methods. Machine Learning provides the dependent variables during the training phase. It learns how to predict the output of unseen data. The system creates a model using labeled data to grasp and understand the feature relations in our dataset. The two main types of supervised learning algorithms are:

Regression: Used to predict continuous values variables (price, salary, age, etc).
Classification: Used to predict or classify the levels of a discrete variable (male or female, true or false, etc).

Unsupervised Learning is a collection of algorithms that learn without any supervision, for example, labeled data, by grouping observations with similar patterns. The two main types of unsupervised learning algorithms are:

Clustering: Used to group or segment datasets with shared attributes in order to extrapolate relationships, it can also be used to detect anomalies.
Principal Component Analysis: Used to project data samples in fewer dimensions and perform dimensionality reduction.

Reinforcement Learning is a feedback-based learning method. The agent learns automatically through positive and negative feedback and improves its performance. In Reinforcement learning, the agent interacts with the environment and explores it.

How does Machine Learning work in Devo?

It is possible to use Machine Learning in Devo thanks to some engines that allow you to train the data and build models so you can upload them to the Devo Platform. See the diagram below explaining the process:

Engines supported by Devo

Here you will find a list of the engines supported by Devo and their versions:

Engine	Version

Engine	Version
H2O	`3.34.0.2`
BigML	`1.8.13`
Catboost	`1.0.4`
ONNX	`1.8.1`

Upload a model in Model Management

Follow these steps to upload a model in Model Management.

Working with models

There are two ways you can work with models, one is through Flow and the other one is through Activeboards. Here is an example of a model to download and an explanation of each case so you can test it.

The example shows if a domain in Devo is legitimate or not. The data has been taken from table demo.ecommerce.data.

Make sure your fields match

Make sure that the fields of your model and when using mlevamodel operation in Activeboards or the ML Single Model Evaluator processor in Flow are the same ones. Otherwise, it will not work.