https://github.com/engineering-research-and-development/demeter-estimate-traceability
DEMETER project: machine learning module for milk quality prediction
https://github.com/engineering-research-and-development/demeter-estimate-traceability
Last synced: 2 months ago
JSON representation
DEMETER project: machine learning module for milk quality prediction
- Host: GitHub
- URL: https://github.com/engineering-research-and-development/demeter-estimate-traceability
- Owner: Engineering-Research-and-Development
- License: other
- Created: 2020-07-06T07:30:57.000Z (almost 5 years ago)
- Default Branch: master
- Last Pushed: 2023-10-03T13:48:59.000Z (over 1 year ago)
- Last Synced: 2025-01-27T21:25:06.343Z (4 months ago)
- Size: 19.1 MB
- Stars: 0
- Watchers: 3
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Estimate milk quality module
Machine learning module for traceability estimation using _**Random Forest**_ algorithm.
The values to predict are a discrete class label such as _**High**_, _**Medium**_ or _**Low**_ milk quality.
During training, it is necessary to provide to the model any historical data that is relevant
to the problem domain and the true value we want the model learns to predict.
The model learns any relationships between the data and the values we want to predict.
The decision tree forms a structure, calculating the best questions to ask to make the most accurate estimates possible.
## Table of contents
* [**Technologies**](#technologies)
* [**Requirements**](#requirements)
* [**Setup**](#setup)
* [**Features**](#features)
* [**Endpoints**](#endpoints)
* [**How to use**](#how-to-use)
* [**Important Notes**](#important-notes)
* [**Troubleshoot**](#troubleshoot)
* [**Developers**](#developers)
* [**Status**](#status)
* [**License**](#license)
## Technologies
| Description | Language | Version |
| :---------------------------------------------- | :------: | :--------------: |
| [Java SE Development Kit 8][1] | Java | 1.8.0_202 |
| [Python 3][2] | Python | 3.7.3 |
| [Apache Maven 3][3] | | 3.6.3 |
| [Apache Tomcat 7][4] | | 7.0.104 |
| [Scikit-learn API][5] | Python | 0.23.2 |
| [JPY API][6] | Java | 0.9.0 |
| [RESTEasy API][7] | Java | 3.12.1.Final |
| [Spring Framework][8] | Java | 4.3.3.RELEASE |
| [Json][9] | | 20200518 |
| [Log4j][10] | | 2.16.0 |
| [Eclipse IDE for Enterprise Java Developers][11] | Java | 2020-06 (4.16.0) |
| [PyDev Python IDE for Eclipse][12] | Python | 7.6.0 |
| [Docker Desktop Community][13] | | 2.3.0.4 (46911) |
[1]: https://www.oracle.com/it/java/technologies/javase/javase-jdk8-downloads.html
[2]: https://www.python.org/downloads/release/python-383/
[3]: http://maven.apache.org/
[4]: https://tomcat.apache.org/download-70.cgi#7.0.104
[5]: https://scikit-learn.org/stable/index.html
[6]: https://jpy.readthedocs.io/en/latest/intro.html
[7]: https://resteasy.github.io/
[8]: https://spring.io/projects/spring-framework
[9]: http://www.JSON.org/
[10]: https://logging.apache.org/log4j/2.x/
[11]: https://www.eclipse.org/downloads/
[12]: http://www.pydev.org/
[13]: https://www.docker.com/products/docker-desktop
## Requirements
* Docker
* Java 1.8
* Python 3
* Maven
* Tomcat 7
The following components are needed for start using this component:
| Components required | Description |
| :-------------------------------- | :------------ |
| [pilot4.2-traslator:latest][14] | AIM traslator |
[14]: https://hub.docker.com/r/demeterengteam/pilot4.2-traslator
## Setup
### Pull the image
`docker pull demeterengteam/estimate-milk-quality:latest`
### Run the application
It's possible to run the application using `docker-compose`
#### Docker-compose
Create a **docker-compose.yml** file into a folder.
*docker-compose.yml content:*
```
version: '3'
services:
milkquality:
image: demeterengteam/estimate-milk-quality:latest
ports:
- '${HOST_PORT}:8080'
environment:
- MQ_AIM_TRASLATOR_PREDICTION_URL=http://${HOST_IP}:${TRASLATOR_SERVICE_HOST_PORT}/demeter-csvManager/rest/traslator/v1/MilkQualityPrediction
- MQ_AIM_TRASLATOR_TRAINING_URL=http://${HOST_IP}:${TRASLATOR_SERVICE_HOST_PORT}/demeter-csvManager/rest/traslator/v1/MilkQualityTraining
```
Create a **.env** file into the same folder of the above docker-compose:
*.env content:*
```
HOST_PORT=9280
HOST_IP=**INSERT HOST IP**
TRASLATOR_SERVICE_HOST_PORT=**INSERT THE PORT OF THE TRASLATOR COMPONENT*
```
Set **HOST_PORT** value to the preferred port to use instead of **9280**.
Set **HOST_IP** value with the correct ip of the host machine.
Set **TRASLATOR_SERVICE_HOST_PORT** value equal to the port used for the traslator service component.
First run the command `docker-compose up` only for the very first time.
Then simply run `docker-compose start` to start the application and `docker-compose stop` to stop it.
Before doing any request, you must upload the data as csv file using the [traslator service][14].
Once started open any REST client (i.e. Postman) and send requests to the application endpoints.
## Features
* **Algorithm Training, Testing and Metrics calculation:**
Receives a dataset of _training features_ as input to perfom the training, test and metrics calculations.
Return an object with the _training result_ that will show all the test records used after the training with the predictions made by the algorithm and the metrics.
* **Predict health condition:**
Receive a dataset of _prediction features_ as input to perform predictions and return an object with the _prediction result_.
## Endpoints
The base URL is composed like:
`http://[HOST]:[HOST_PORT]/EstimateMilkQualityModule/ENDPOINT`
Headers settings:
| Key | Value |
| :----------- | :--------------- |
| Content-Type | application/json |
| Accept | application/json |
Endpoint informations:
| URL | Type | Used for | Input | Output |
| :-------------------------- | :-----: | :------------------------------------------------------------------ | :---- | :--------------------------------------- |
| `/v1/milkQualityTraininig` | **GET** | Train the algorithm, calculate the metrics and send the result data | | AIM Json output data result with metrics |
| `/v1/milkQualityPrediction` | **GET** | Estimate the health condition and send the result data | | AIM Json output data result |
## How to use
**TO DO**
## Important Notes
The application image don't contains any training data model preloaded, so the first request to execute
must be a training one. That will create the models needed for the prediction tasks.
Isn't necessary to execute another training on next use of the application except for the purpose of improve training accuracy.
## Troubleshoot
### Errors
The service provide error messages as response and write the details into logs.
There are two different error codes that are useful to identify where the error occured:
* **Code: 1** - The error occured within the java classes
* **Code: 2** - The error occured within the python modules
## Developers
* [Luigi di Corrado](https://github.com/luidicorra)
## Status
Project is: _in progress_
## License
[](https://www.gnu.org/licenses/agpl-3.0)