Student Performance Prediction Project Report

1. Introduction

This project focuses on predicting student GPA using machine learning techniques. The main
objective is to analyze student performance data, clean and preprocess the dataset, identify
relationships between variables, and build a predictive model capable of estimating GPA values.

The project also prepares the dataset for visualization and dashboard creation in Power BI by
adding a new column containing predicted GPA values.

2. Project Objectives

The main objectives of this project are:

 Analyze student performance data.

 Clean and preprocess the dataset.

 Handle missing values and outliers.

 Explore relationships between variables.

 Build a machine learning model to predict GPA.

 Evaluate the model using regression metrics.

 Generate a final dataset containing predicted GPA values for Power BI visualization.

3. Dataset Description

The dataset contains student-related academic information and performance indicators used to
predict GPA.

Main Characteristics of the Dataset

 Numerical student performance features.

 Target variable: GPA.

 Some missing or invalid values.

 Duplicate records.

 Possible outliers affecting model accuracy.

4. Methodology

The methodology followed in this project consists of several stages:

1. Data Collection

2. Data Cleaning and Preprocessing

3. Exploratory Data Analysis (EDA)

4. Feature Selection

5. Data Splitting

6. Model Training

7. Model Evaluation

8. Prediction Generation

9. Exporting Results for Power BI

5. Data Cleaning and Preprocessing

Data preprocessing is an important step in machine learning because raw datasets often contain
missing values, duplicates, and inconsistent data.

5.1 Importing Libraries

Several Python libraries were used in this project:

 Pandas → data manipulation.

 NumPy → numerical operations.

 Matplotlib and Seaborn → data visualization.

 Scikit-learn → machine learning and evaluation.

5.2 Loading the Dataset

The dataset was loaded from an Excel file using Pandas.

pd.read_excel(file_path)
5.3 Handling Missing Values

The dataset contained missing or invalid values. These values were converted into NaN values
and replaced using the mean strategy.

SimpleImputer(strategy='mean')

Using the mean helps preserve the dataset size while reducing information loss.

5.4 Removing Duplicate Rows

Duplicate rows were identified and removed to improve data quality.

df.drop_duplicates()

5.5 Removing Unnecessary Columns

The StudentID column was removed because it does not contribute to GPA prediction.

5.6 Outlier Detection and Removal

The Interquartile Range (IQR) method was used to detect and remove outliers.

IQR Formula

 IQR = Q3 − Q1

 Lower Bound = Q1 − 1.5 × IQR

 Upper Bound = Q3 + 1.5 × IQR

Rows containing values outside the acceptable range were removed.

This step helps improve model stability and prediction accuracy.

6. Exploratory Data Analysis (EDA)

Exploratory Data Analysis was performed to better understand the dataset and relationships
between variables.

6.1 Correlation Heatmap

A correlation heatmap was created using Seaborn.

[Link](df[numeric_cols].corr(), annot=True, cmap='coolwarm')

The heatmap helps identify:

 Strong positive correlations.

 Features highly related to GPA.

This analysis is important for understanding how different variables influence academic
performance.

6.2 Actual vs Predicted GPA Visualization

A scatter plot was generated to compare actual GPA values with predicted GPA values.

This visualization helps evaluate how close the model predictions are to real values.

7. Feature Selection

The target variable selected for prediction was:

 GPA

All remaining columns were used as input features.

X = [Link]('GPA', axis=1)

y = df['GPA']

8. Modeling Process

8.1 Train-Test Split

The dataset was divided into training and testing sets.

 80% Training Data

 20% Testing Data

train_test_split(test_size=0.2, random_state=42)

The training set was used to train the model, while the testing set was used to evaluate model
performance.

8.2 Linear Regression Model

A Linear Regression model was selected for predicting GPA.

Why Linear Regression?

 Simple and easy to interpret.

 Effective for numerical prediction tasks.

 Suitable for identifying relationships between variables.

 Computationally efficient.

Model Training

The model was trained using:

model = LinearRegression()

[Link](X_train, y_train)

The model learns relationships between student features and GPA.

9. Model Evaluation

Model performance was evaluated using regression metrics.

9.1 Mean Squared Error (MSE)

Mean Squared Error measures the average squared difference between actual and predicted
values.

mean_squared_error(y_test, predictions)

Lower MSE values indicate better model performance.

And it was 0.101

9.2 R² Score

The R² Score measures how well the model explains variance in GPA.

r2_score(y_test, predictions)

Interpretation

 R² close to 1 → strong prediction performance.

 R² close to 0 → weak prediction performance.

10. Prediction Generation

After training the model, GPA predictions were generated for all rows in the dataset.

df['Predicted_GPA'] = [Link](X)

A new column called Predicted_GPA was added.

This allows comparison between:

 Actual GPA

 Predicted GPA

The updated dataset was exported to Excel for Power BI dashboard creation.

df.to_excel('Student_performance_with_predictions.xlsx', index=False)

11. Power BI Integration

The final dataset containing predicted GPA values can be imported into Power BI to create:

 KPI dashboards

 Comparison charts

 Correlation visuals

 GPA distribution analysis

 Actual vs Predicted GPA comparisons

 Student performance insights

This improves data visualization and decision-making.

12. Results and Findings

The project successfully:

 Cleaned and processed the dataset.

 Removed duplicates and outliers.

 Identified correlations between variables.

 Built a Linear Regression prediction model.

 Generated GPA predictions.

 Prepared a dataset for Power BI visualization.

The Linear Regression model provided a simple and effective approach for GPA prediction.

13. Challenges Encountered

Some challenges during the project included:

 Missing and inconsistent values.

 Presence of outliers.

 Data quality issues.

 Ensuring the model generalized well on testing data.

These challenges were addressed through preprocessing and proper evaluation methods.

14. Conclusion

This project demonstrates the complete machine learning workflow for predicting student GPA.

The process included data cleaning, preprocessing, exploratory analysis, model training,
evaluation, and prediction generation.

Linear Regression successfully predicted GPA values and produced a final dataset suitable for
Power BI visualization and analysis.

The project highlights how machine learning and data analytics can support educational
performance analysis and decision-making.