How To Use

Brief description:

Automatic predictive model constructor (APMC) allows to train and develop supervised machine learning models. Models are training based on data-set, which includes a set of predictors (independent variables) and one target value (dependent variable). When a model is trained allows predicting so far unknown target value based on any predictors.

More information and examples data sets in APMC documentation.

Example data sets:

for classification purposes:

--> Iris

for regression purposes:

--> USA Housing

How to use step-by-step:

1. Prepare your data

Data must be prepared in accordance with the following rules:

-> data set must be with “.csv”, ".xls" or ".xlsx" extensions.

-> default delimiter type is “,” (in case of "csv" format)!

-> in case of float numbers use dots instead of commas

-> predictors are in columns

-> every column need a label name

-> last column is target values

NOTE 1: Maximum number of predictors is 10, maximum number of records is 10000.

NOTE 2: In case of classification models target value must be a string, in case of regression models target value must be a number.  

2. Project name

Type name of your project, your project, model and a statistical report will be saved with this name

3. Select type of model:

APMC support classification and regression algorithms. If target values are quantitative variables (price, weight, height) select a regression model. If target value is a categorical variable (gender, species, blood type) select classification model.

4. Normalization

Normalization may have a positive influence on model training, this function is optional.

5. Pre-Training

When the user finishes above-mentioned steps and click "Pre-train" button.

6. Select model

After step no. 5, set of models are ready, the user needs to select which one will move on to the next stage of training. We suggest which algorithm probably is [for current data set] the best, based on the model`s metrics.

7. Prediction

After click the "TRAIN" button, the model is prepared for use, when it is ready, the "PREDICT" button redirects to the page with the user's model collection.

8. Models collection

After the redirect, the user has a preview of the model's metrics, and the ability to predict target values.

9. Statistical report and tree graph

To get the statistical report, go to the "model" tab in the user panel, there is the download report button. If the selected algorithm is: random forest regression or random forest classification, it is also possible to download a tree graph.