Last update: 14-02-2024

The classification predicament involves assigning labels or categories to data instances based on observed features. Consider, for instance, the task of discriminating between “spam” and “non-spam” emails. This constitutes a classification task, where the algorithm must acquire the ability to discern patterns that distinguish the two email types based on their keywords, structure, or other attributes. Classification algorithms employ a training dataset containing pre-labeled examples to learn these patterns. When faced with unclassified data, the algorithm applies the acquired patterns to predict the class to which they belong, enabling efficient and precise automation in the categorization of new cases. Algorithms like those in FuzzyClass address this task by leveraging data probabilities and characteristics, thus becoming valuable tools for addressing intricate and ambiguous classification problems.

A package manual that showcases the existing classifiers and demonstrates how to use it can be found at the following link: https://cran.r-project.org/package=FuzzyClass/FuzzyClass.pdf

Below is the list of packages on which FuzzyClass depends. However, during its installation, FuzzyClass automatically installs the dependencies:

• caTools
• doParallel
• e1071
• EnvStats
• foreach
• MASS
• maxLik
• mvtnorm
• purrr
• dplyr
• Rdpack
• rootSolve

# Installation
install.packages("devtools")
devtools::install_github("leapigufpb/FuzzyClass")

Once installed, you can load the FuzzyClass package into your R session:

# Package import
library(FuzzyClass)

To demonstrate the usage of FuzzyClass, let’s look at reading and preparing data:

library(FuzzyClass)

#' ---------------------------------------------
#' The following shows how the functions are used:
#' --------------
#' Reading a database:
#'
#' Actual training data:
data(VirtualRealityData)

VirtualRealityData <- as.data.frame(VirtualRealityData)

# Splitting into Training and Testing
split <- caTools::sample.split(t(VirtualRealityData[,1]), SplitRatio = 0.7)
Train <- subset(VirtualRealityData, split == "TRUE")
Test <- subset(VirtualRealityData, split == "FALSE")
# ----------------

test = Test[,-4]

Let’s delve into the example of using the Fuzzy Gaussian Naive Bayes algorithm with fuzzy parameters:

# --------------------------------------------------
# Fuzzy Gaussian Naive Bayes with Fuzzy Parameters


fit_FGNB <- GauNBFuzzyParam(train =  Train[,-4],
                                    cl = Train[,4], metd = 2, cores = 1)


print(fit_FGNB)
#> 
#> Fuzzy Gaussian Naive Bayes Classifier for Discrete Predictors
#> 
#> Variables:
#> [1] "V1" "V2" "V3"
#> Class:
#> [1] "1" "2" "3"
saida <- predict(fit_FGNB, test)
Table <- table(factor(Test[,4]), saida)
Table
#>    saida
#>      1  2  3
#>   1 50  5  2
#>   2  6 42 15
#>   3  0 11 49

#Accuracy:
sum(diag(Table))/sum(Table)
#> [1] 0.7833333

saidaMatrix <- predict(fit_FGNB, test, type = "matrix")

Additionally, you can visualize the results:

# --------------------------------------------------
# head view

saida |> head()
#> [1] 1 1 1 1 1 1
#> Levels: 1 2 3

saidaMatrix |> head()
#>              1           2            3
#> [1,] 0.9989435 0.001056437 9.262171e-08
#> [2,] 0.9939728 0.006011989 1.523144e-05
#> [3,] 0.8213097 0.116368282 6.232206e-02
#> [4,] 0.9946096 0.005386036 4.371040e-06
#> [5,] 0.8684685 0.069905455 6.162602e-02
#> [6,] 0.8015720 0.145765858 5.266218e-02

This code appears to be related to the application of a classification algorithm called “Fuzzy Gaussian Naive Bayes with Fuzzy Parameters.” An analysis of the steps present in the code:

1. Model Training (fit_FGNB):
   • A Fuzzy Gaussian Naive Bayes model is being fitted to the training data.
   • The training set consists of attributes (Train[,-4]) and classes (Train[,4]), where the categorical response variable or label is in column 4.
2. Prediction and Confusion Matrix Creation:
   • The predict function is used to make predictions based on the fitted model using the test set (test).
   • A confusion matrix (Table) is created using the table function. The confusion matrix compares the actual (expected) classes with the classes predicted by the model.
3. Accuracy Calculation:
   • The accuracy of the model is calculated by dividing the sum of the diagonal values of the confusion matrix (true positives and true negatives) by the total sum of the confusion matrix. This provides a measure of how well the model is performing predictions.

Overall, this code performs the training of a Fuzzy Gaussian Naive Bayes model with fuzzy parameters, makes predictions using the test set, creates a confusion matrix to evaluate the model’s performance, and calculates its accuracy.

This enhanced documentation provides a comprehensive guide to using the FuzzyClass package for probabilistic classification tasks. It covers installation, package usage, data preparation, and examples of applying the Fuzzy Gaussian Naive Bayes algorithm with fuzzy parameters. Feel free to explore the package further to leverage its capabilities for your classification tasks.

If you would like to contribute to FuzzyClass, please follow these steps:

1. Fork the FuzzyClass repository on GitHub.
2. Create a new branch for your contribution.
3. Make your changes to the code or documentation.
4. Test your changes thoroughly.
5. Add or update documentation for your changes.
6. Submit a pull request to the main FuzzyClass repository.
7. The FuzzyClass maintainers will review your pull request and may ask you to make some changes before it is merged. Once your pull request is merged, your contribution will be available to all FuzzyClass users.

• Please use descriptive commit messages that explain what your changes do.
• If you are making a large or complex change, please consider creating an issue in the FuzzyClass repository first to discuss your plans with the maintainers.
• Please be patient while your pull request is reviewed. The maintainers may be busy with other things, but they will get to your pull request as soon as they can.

If you find a bug in FuzzyClass, please report it by creating an issue on the FuzzyClass repository on GitHub at the link: https://github.com/leapigufpb/FuzzyClass/issues. When reporting an issue, please include the following information:

1. A clear and concise description of the bug.
2. The steps to reproduce the bug.
3. The expected behavior.
4. The actual behavior.
5. Any relevant screenshots or code snippets.
6. If possible, please also include the version of FuzzyClass that you are using.

The FuzzyClass maintainers will review your issue and may ask you for more information before they can fix the bug. Once the bug is fixed, a new release of FuzzyClass will be made available.

Here are some additional tips for reporting issues to FuzzyClass:

• Please be as specific as possible when describing the bug.
• If you can, try to reproduce the bug on a clean installation of R.
• Include the output of sessionInfo() when reporting a bug. This will help the maintainers to diagnose the problem.
• Please be patient while your issue is being reviewed. The maintainers may be busy with other things, but they will get to your issue as soon as they can.

I hope this helps! Let me know if you have any other questions.