Getting started as a model developer¶

The Chap platform brings several advantages to you as a model developer:

It provides a broad range of supporting functionality for modelling, allowing you to focus on what is unique in your project, while relying on Chap for data parsing, model tuning, rigorous model assessment and more.
It allows you to evaluate your model side-by-side with a range of other models in the field
It comes with native connection to DHIS2 and the DHIS2 Modelling App, and includes mechanisms to disseminate your model to a large number of health ministries that use DHIS2 as their national health information system.

How a model runs in Chap¶

A Chap model is a train script and a predict script, in R or Python, plus a short MLproject file that tells Chap how to run them. Once your model is Chap-compatible, you can use it in two ways:

On your own machine, with the chap command-line tool. One command evaluates the model on a dataset and plots its forecasts:
```
chap eval --model-name . --dataset-csv data.csv --output-file eval.nc --plot --run-config.as-chapkit
```
In the Chap Modeling App, where users can evaluate it on DHIS2 data, compare it with other models and make forecasts. For this, you package the model as a Docker image and add it to a Chap installation, see Running Your Own Model.

Where to start¶

First set up your machine. Then pick the guide that fits you:

You...	Start here
already have a forecasting model in R or Python	Bring your existing model to Chap
want to build a new model	Build a new model from a template
are not used to the terminal, git, or installing Python or R packages	Development tools, then come back here
want to learn statistical modelling and forecasting	Learning modelling

Both model guides use the same MLproject format and the same chap eval command. They only differ in where you start: from your own code, or from a template.

Strictly, Chap only requires that a model exposes the chapkit REST API. The train and predict scripts and the MLproject file are the recommended way to get there, since chapkit then provides the API for you. Advanced users can write a service that implements the API directly, see the chapkit documentation. Most models do not need this.