README.md

In ginozab/mspredictr:

mspredictr

mspredictr is an R package that helps ease the process of running data mining and supervised machine learning to make predictions on target data.

mspredictr makes use of the caret package in the R language for statistical computation to run the machine learning and data mining techniques.

Functions available to call

There are three different categories of functions provided by the mspredictr package. These categories include prediction functions, benchmarking functions, and visualization functions.

| Function Type | Function Name | |---------------|------------------| | Prediction | cat_prediction | | Prediction | num_predict | | Prediction | cat_majority | | Prediction | reg_average_pred | | Benchmarking | benchmark | | Benchmarking | pred_bench | | Benchmarking | pred_frame | | Visualization | graph_preds | | Visualization | graph_overall |

Prediction Functions

There are two main prediction functions made available in the mspredictr package along with two base prediction functions. The first prediction function is the cat_prediction function which helps facilitate making predictions on categorical data. The second function is the num_predict function which helps make predictions on numerical data types.

These two functions take require three parameters. The first parameter is the dataset from which training data and target prediction data. The second is the machine learning algorithm that should be used to make the predictive model for making predictions. The final parameter is to specify on which data the function should be making predictions.

The two base functions are targeted for two different target predictor data types. The first base prediction function is targeted for categorical data predictions, and predicts the majority of a categorical occurrence in the training data. This function is the cat_majority function. The second base function is target for numerical base predictions, and this function predicts the average of the numerical occurrences in training data. This function is the reg_average_pred function.

Benchmarking Functions

The main benchmarking function is the benchmark function which can help automate run however many desired trials of predictions using a list of algorithms and a list of target prediction data. This benchmark function makes use of the next two functions available in the package, the pred_bench function and the pred_frame function. The pred_bench function automates a desired number of trials using a list of methods to predict on one target data. The pred_frame function is a frame for all possible prediction configurations. The pred_frame function can run categorical and numerical predictions using any type of method to predict any desired target data.

Visualization Functions

There are two visualization functions available in the mspredictr package. The graph_preds and graph_overall functions take in the data frame returned from a successful run of the benchmark function. These two functions then create a box and whisker plot of he results. graph_preds makes a faceted box and whisker that is split based on the values in the program column. graph_overall creates a single box and whisker plot that visualizes the prediction results from all of the different programs.

Data Available for Use

Below is a table of all the available data frames in the mspredictr package.

| Data Frame | Target Data Type | Description | |----------------|------------------|-----------------------------------------| | totalCDCatCP | Categorical | Three Category Codepro Mutation Scores | | totalCDCatEvo | Categorical | Three Category Evosuite Mutation Scores | | totalCDCatMan | Categorical | Three Category Manual Mutation Scores | | totalCDCatQCP | Categorical | Six Category Codepro Mutation Scores | | totalCDCatQEvo | Categorical | Six Category Evosuite Mutation Scores | | totalCDCatQMan | Categorical | Six Category Manual Mutation Scores | | totalCDRegCP | Numerical | Numerical Codepro Mutation Scores | | totalCDRegEvo | Numerical | Numerical Evosuite Mutation Scores | | totalCDRegMan | Numerical | Numerical Manual Mutation Scores |

Usage Example

Here is a link to an asciinema video example of how to use the mspredictr package for making predictions using data mining and supervised machine learning techniques.

ginozab/mspredictr documentation built on May 17, 2019, 4:20 a.m.