Firstly you have to import a package called "classification" that contains all important functions for classifying a
dataset consisting of float values:
import classification.ClassificationOfFloatValues;
The next step is to create an object for this classification (ob is used as a default name for an object):
ClassificationOfFloatValues ob = new ClassificationOfFloatValues(dataset);
The dataset variable should contain the name of the dataset that should be classified as a string.
The dataset has to be in the same folder as the main file.
If the dataset has an index or a header (or both), it has to be indecaded by the user.
If there is a header you have to call ob.setIndex(true); or/and ob.setHeader(true);.
The default value for these is false because it is expected that the dataset does not have an index or header.
Most datasets do have a header and an index so make sure, if your dataset has a header or an index, to include this part in your program.
The following functions are required for classifying the data.
Firstly you have to call ob.dataProcessing();
ob.dataSubdivision();
ob.distanceClassification();
For evaluating the predicted results you can call ob.evaluateResults();.
There are multiple ways to show how the results should be displayed.
The ob.setEvaluation(model) functions sets the evaluation models which are going to be calculated and printed.
model should contain one of the names below as a string.
Confusion Matrix: Printing a normal confusion matrix for every class (size: class x class).
Simple Confusion Matrix: Printing a simplified confusion matrix for every class with true positives and false positives (size: class x 2).
NormalizedConfusion Matrix: Printing a normalized confusion matrix with the format of the confusion matrix as explained
above. The values that are displayed a normalized (values between 0 and 1).
There is a script that explains the programs function and also explains the data manipulation in detail.
You can find the description here.
If you need help if applying the algorithm to your projects, feel free to ask.
Contributors names and contact info
- Max Wenk
The current built is v-0.1.
It is possible to classify a dataset which contains only float values.
It is important to consider that the weight for every class is the same.
This project is licensed under the "GNU Affero General Public License v3.0" License - see the LICENSE.md file for details.