app.py

import gradio as gr
from tensorflow.keras.models import load_model
from tensorflow.keras.metrics import MeanAbsoluteError
import tensorflow as tf
import numpy as np

# Register custom metric
tf.keras.utils.get_custom_objects().update({"mae": MeanAbsoluteError()})

# Load your trained model
model = load_model("student_performance_model.h5")

# Define prediction function
def predict_performance(Gender, AttendanceRate, StudyHoursPerWeek, PreviousGrade, ExtracurricularActivities, ParentalSupport):
    # Encoding Gender (Male/Female)
    Gender_Male = 1 if Gender == 'Male' else 0
    Gender_Female = 1 if Gender == 'Female' else 0
    
    # One-hot encode ParentalSupport (Low/Medium/High)
    ParentalSupport_Low = 1 if ParentalSupport == 'Low' else 0
    ParentalSupport_Medium = 1 if ParentalSupport == 'Medium' else 0
    ParentalSupport_High = 1 if ParentalSupport == 'High' else 0
    
    # ExtracurricularActivities is now numeric (0-3)
    # No transformation needed, it's a numeric input already
    
     # Prepare input array as a NumPy array
    input_data = np.array([
        [Gender_Male, Gender_Female, AttendanceRate, StudyHoursPerWeek, PreviousGrade, ExtracurricularActivities, 
         ParentalSupport_Low, ParentalSupport_Medium, ParentalSupport_High]
    ])
    
    # Predict the student's performance
    prediction = model.predict(input_data)
    return prediction[0]

# Gradio interface
interface = gr.Interface(
    fn=predict_performance,
    inputs=[
        gr.Dropdown(choices=['Male', 'Female'], label='Gender'),
        gr.Number(label='Attendance Rate (%)'),
        gr.Number(label='Study Hours Per Week'),
        gr.Number(label='Previous Grade'),
        gr.Slider(0, 3, step=1, label='Number of Extracurricular Activities'),  # Updated: Numeric slider (0-3)
        gr.Dropdown(choices=['Low', 'Medium', 'High'], label='Parental Support')
    ],
    outputs="text"
)

# Launch the interface
interface.launch()