Building machine learning models should be easy! (you see too many steps below to start? come on, it is only like that at the beginning, go and do it! Trust me it is easy!)
- Login and create a project.
- Go into a project by clicking on its title.
- In project main view, you should be redirected to
Sourcesif it is an empty project.
Sourcesview, please add a dataset by clicking
Add new dataset.
Add new datasetbox please provide a name for your dataset and select a data file. (Please check available data formats)
- OK, after uploading dataset please wait for a while, we are computing a dataset basic statistics. You will see a message in data source that it is ready for preview in MLJAR service.
- Please click on
Previewor go to
Previewfrom left menu and select your dataset. Each dataset should be accepted before analysis. You need to select attributes which will be used as models input (select for them
Use itvalue) and you need to specify
Targetattribute, which will be model output. For attributes not used in analysis please set column usage to
Don't use. You can also specify
Idcolumn if exists. After selecting attributes please
Acceptdataset in top of the view.
- With accpeted dataset you are ready to define machine learning experiments (yes, you have to experiment a lot if you want to do good ML :) ). Please go to
Computeview, please click on
Add new experiment.
While defining a new experiment you need to define:
- input dataset
- learning algoritms
- tuning parameters
Saveexperiment and click on it on the list. Check if its details looks like expected and click
Start experimentafter that all the machine learning magic will start!
- To review models go to
Resultsfrom left menu. To check algorithm details click on it.
To get predictions from trained models you can download model in
Deploy view or you can upload test dataset into MLJAR and use it for prediction. This situation is described below.
- Upload you data file in
- Accept attributes usage in
Previewfor uploaded dataset.
- Go to
Predictview, in top right corner select dataset which will be used for prediction.
- Select algorithm that will be used, by selecting checkbox.
Start Predictionand wait a while. Click
Refreshon the bottom of the page to see if prediction is computed.
- Your predictions will be displayed in the bottom of the page where you can download it.
To deploy your model:
- you can download it from
Deployview - we are working on script that will allow you easy deploy models locally, so checkout our github
- or your model can be accessed by REST API (we are working on this feature !!!)