Embedding Machine Learning Model in Shiny Web App

Introduction

In recent years, machine learning has become a crucial aspect of data analysis and visualization. One popular framework for building interactive web applications is Shiny. Shiny allows users to create custom web pages with real-time data updates using R’s powerful data science libraries, including machine learning models. In this article, we will explore how to integrate a machine learning model into a Shiny web app.

Prerequisites

To build an embedded machine learning model in a Shiny web app, you need the following prerequisites:

• Basic knowledge of R programming language and its data structures
• Familiarity with machine learning concepts and techniques (e.g., linear regression, classification)
• Installation of necessary packages such as shiny, dplyr, tidyr, and caret

Loading and Preparing Data

The first step in integrating a machine learning model into a Shiny web app is to load and prepare the necessary data. This includes:

• Loading your dataset (e.g., boxing dataframe)
• Splitting the data into training and testing sets
• Creating feature variables (e.g., weight, height, reach)
• Encoding categorical variables (if any)

# Load necessary libraries
library(shiny)
library(dplyr)
library(tidyr)
library(caret)

# Load the dataset
boxing <- read.csv("data/boxing.csv")

# Split the data into training and testing sets
set.seed(123)  # for reproducibility
trainIndex <- createDataPartition(boxing$opponent, p = .7, list = FALSE)
trainBoxing <- boxing[trainIndex, ]
testBoxing <- boxing[-trainIndex, ]

# Create feature variables
boxing$weight <- as.numeric(boxing$weight)  # convert to numeric
boxing$height <- as.numeric(boxing$height)  # convert to numeric
boxing$reach <- as.numeric(boxing$reach)  # convert to numeric

# Encode categorical variable 'division'
boxing$division <- as.factor(boxing$division)

# Create a testing set for the model
testSet <- boxing[-c(1, 2), ]  # exclude first two rows

Building and Training the Machine Learning Model

Once your data is prepared, you can build and train your machine learning model. This step involves:

• Choosing an appropriate algorithm (e.g., linear regression, classification)
• Tuning hyperparameters using techniques such as grid search or cross-validation
• Evaluating the model’s performance on a test set

# Build the model
model <- lm(outcome ~ weight + height + reach + division, data = trainBoxing)

# Evaluate the model's performance
summary(model)  # see coefficients and R-squared value

Integrating the Machine Learning Model into Shiny

To integrate your machine learning model into a Shiny web app, you need to:

• Define user input variables (e.g., input$names, input$names2)
• Create an output variable that uses the trained model to make predictions
• Use R’s powerful data visualization libraries (e.g., ggplot2) for interactive plots

# Define UI
ui <- fluidPage(
    # User input variables
    numericInput("names", "Name:", value = 0),
    selectInput("division", "Division:", levels(trainBoxing$division)),
    
    # Action button to trigger prediction
    actionButton("gobutton", "Predict"),
    
    # Output variable that uses the model for prediction
    dataTableOutput("predictions")
)

# Define server function
server <- function(input, output) {
    # Create a reactive value that holds the trained model
    model <- reactive({
        cat_model <- read_pickle("catmodel.pkl")  # load trained model
        return(cat_model)
    })

    # Create an output variable that uses the trained model for prediction
    output$predictions <- renderDataTable({
        # Get user input values
        namesInput <- input$names
        divisionSelected <- input$division
        
        # Make prediction using the trained model
        prediction <- model()$predict(namesInput, namesInput, type = "response")
        
        # Create a data frame for output
        dataFrame(
            outcome = ifelse(prediction == 1, "Win", 
                              ifelse(prediction == -1, "Loss", "Draw"))
        )
    })
}

Interacting with the Shiny Web App

Once you have integrated your machine learning model into a Shiny web app, you can interact with it by:

• Selecting input values (e.g., names, division)
• Clicking the “Predict” button to trigger prediction
• Viewing the output data frame in the table

Conclusion

In this article, we explored how to integrate a machine learning model into a Shiny web app. We discussed loading and preparing data, building and training the model, integrating it into Shiny, and interacting with the web app.

With these steps, you can create an interactive web application that leverages your machine learning skills to make predictions on real-time data.

Additional Tips

• Always ensure that your machine learning models are robust and reliable.
• Continuously monitor and update your model’s performance on a test set.
• Use techniques such as cross-validation and grid search to tune hyperparameters for better results.

Last modified on 2025-04-07