MISCONCEPTION / app.py
harrycool12's picture
Update app.py
a66ea0f verified
Raw History Blame Contribute Delete
13.3 kB
import streamlit as st
import pandas as pd
import numpy as np
from preprocessing import TextPreprocessor
from model import MisconceptionClassifier
from visualization import ResultVisualizer
from sklearn.model_selection import train_test_split
# Page config
st.set_page_config(
page_title="Math Misconception Analyzer",
layout="wide"
)
# Initialize components
@st.cache_resource
def load_preprocessor():
return TextPreprocessor()
@st.cache_resource
def load_model():
return MisconceptionClassifier.load_model()
preprocessor = load_preprocessor()
model = load_model()
visualizer = ResultVisualizer()
# Main interface
st.title("Mathematical Misconception Analyzer")
st.write("""
This tool analyzes student responses to identify mathematical misconceptions using advanced NLP techniques.
Upload your data to get detailed insights about common misconceptions and patterns.
""")
# Tabs for different functionalities
tab1, tab2, tab3 = st.tabs(["Analyze New Data", "Model Retraining", "Quick Analysis"])
with tab1:
# File upload for analysis
uploaded_file = st.file_uploader("Upload CSV file for analysis", type=['csv'], key='analysis_upload')
if uploaded_file is not None:
try:
# Load and display data
df = pd.read_csv(uploaded_file)
st.subheader("Data Preview")
st.dataframe(df.head())
# Column selection
text_column = st.selectbox(
"Select the column containing student responses",
df.columns
)
label_column = st.selectbox(
"Select the column containing misconception labels (if available)",
['None'] + list(df.columns)
)
if st.button("Analyze"):
with st.spinner("Processing data..."):
# Extract features with advanced NLP
features = preprocessor.extract_features(df, text_column)
# Show feature analysis
st.subheader("Text Analysis")
col1, col2, col3 = st.columns(3)
with col1:
st.metric("Average Response Length",
f"{features['text_length'].mean():.1f} words")
with col2:
st.metric("Mathematical Terms Density",
f"{features['math_density'].mean():.2f}")
with col3:
st.metric("Total Responses", len(df))
# Show topic modeling results if available
if hasattr(preprocessor, 'topic_keywords'):
st.subheader("Topic Analysis")
topic_fig = visualizer.plot_topic_distribution(
features,
preprocessor.topic_keywords
)
if topic_fig:
st.plotly_chart(topic_fig)
# Train model if labels are available
if label_column != 'None':
# Check if we have enough data
if len(df) < 4:
st.warning("Dataset is too small for reliable model training. Please provide more samples.")
st.stop()
X_train, X_test, y_train, y_test = train_test_split(
df[text_column],
df[label_column],
test_size=0.2,
random_state=42
)
# Train and evaluate model
with st.spinner("Training model..."):
model.train(X_train, y_train)
predictions = model.predict(X_test)
probabilities = model.predict_proba(X_test)
metrics = model.get_model_metrics()
# Display results
st.subheader("Model Performance")
# Show cross-validation results if available
cv_results = visualizer.plot_model_metrics(metrics)
if cv_results is not None:
if isinstance(cv_results, tuple):
for fig in cv_results:
st.plotly_chart(fig)
else:
st.plotly_chart(cv_results)
# Show prediction distribution and confidence
col1, col2 = st.columns(2)
with col1:
st.plotly_chart(
visualizer.plot_misconception_distribution(
pd.Series(predictions)
)
)
with col2:
st.plotly_chart(
visualizer.plot_confidence_scores(
probabilities,
model.pipeline.classes_
)
)
# Show embedding clusters
st.subheader("Response Embedding Analysis")
embeddings = np.array([preprocessor.get_document_embedding(text)
for text in X_test])
embedding_fig = visualizer.plot_embedding_clusters(
embeddings,
y_test
)
if embedding_fig is not None:
st.plotly_chart(embedding_fig)
else:
st.info("Not enough samples for meaningful embedding visualization. More data is needed for this analysis.")
# Feature importance
st.subheader("Feature Importance Analysis")
st.plotly_chart(
visualizer.plot_feature_importance(
model.get_feature_importance()
)
)
# Show detailed metrics
st.subheader("Classification Report")
report_df = pd.DataFrame(metrics['classification_report']).T
st.dataframe(report_df)
else:
st.warning("No labels provided. Only preprocessing results available.")
st.write("Extracted Features:")
st.dataframe(features)
except Exception as e:
st.error(f"Error processing file: {str(e)}")
with tab2:
st.header("Model Retraining")
st.write("""
Upload additional labeled data to retrain the model and improve its performance.
The model will be updated with the new data while maintaining its existing knowledge.
""")
# Training history
st.subheader("Training History")
history_fig = visualizer.plot_training_history(model.get_training_history())
if history_fig:
st.plotly_chart(history_fig)
else:
st.info("No training history available yet.")
# File upload for retraining
retrain_file = st.file_uploader("Upload CSV file for retraining", type=['csv'], key='retrain_upload')
if retrain_file is not None:
try:
# Load new data
new_data = pd.read_csv(retrain_file)
st.write("New Training Data Preview:")
st.dataframe(new_data.head())
# Column selection for retraining
text_col = st.selectbox(
"Select the column containing student responses",
new_data.columns,
key='retrain_text_col'
)
label_col = st.selectbox(
"Select the column containing misconception labels",
new_data.columns,
key='retrain_label_col'
)
if st.button("Retrain Model"):
with st.spinner("Retraining model with new data..."):
# Split new data
X_new_train, X_new_test, y_new_train, y_new_test = train_test_split(
new_data[text_col],
new_data[label_col],
test_size=0.2,
random_state=42
)
# Retrain model
model.train(X_new_train, y_new_train, is_retraining=True)
# Save updated model
model.save_model()
# Evaluate on new test data
new_predictions = model.predict(X_new_test)
new_probabilities = model.predict_proba(X_new_test)
new_metrics = model.get_model_metrics()
# Show updated performance
st.subheader("Updated Model Performance")
# Show cross-validation results
cv_results = visualizer.plot_model_metrics(new_metrics)
if cv_results is not None:
if isinstance(cv_results, tuple):
for fig in cv_results:
st.plotly_chart(fig)
else:
st.plotly_chart(cv_results)
# Show updated training history
st.subheader("Updated Training History")
updated_history_fig = visualizer.plot_training_history(model.get_training_history())
if updated_history_fig:
st.plotly_chart(updated_history_fig)
st.success("Model successfully retrained and saved!")
except Exception as e:
st.error(f"Error during retraining: {str(e)}")
with tab3:
st.header("Quick Misconception Analysis")
st.write("""
Enter a student's response to analyze potential mathematical misconceptions.
The model will identify the type of misconception and provide confidence scores.
""")
# Single response analysis
student_response = st.text_area(
"Enter student response",
placeholder="Example: When multiplying decimals, more decimal places means bigger number"
)
if st.button("Analyze Response"):
if student_response:
with st.spinner("Analyzing response..."):
# Get prediction and probability
prediction = model.predict([student_response])[0]
probabilities = model.predict_proba([student_response])[0]
# Show results
st.subheader("Analysis Results")
# Show detected misconception with high emphasis
st.markdown(f"""
### Detected Misconception:
**{prediction}**
""")
# Show confidence scores
st.subheader("Confidence Scores")
probs_df = pd.DataFrame({
'Misconception Type': model.pipeline.classes_,
'Confidence': probabilities
}).sort_values('Confidence', ascending=False)
st.dataframe(probs_df)
# Extract features for insight
features = preprocessor.preprocess_text(student_response)
# Show key mathematical elements detected
st.subheader("Key Mathematical Elements Detected")
col1, col2, col3 = st.columns(3)
with col1:
st.metric("Numbers Found", features['numbers_found'])
with col2:
st.metric("Operations Found", features['operations_found'])
with col3:
st.metric("Math Terms", features['math_terms_found'])
else:
st.warning("Please enter a student response to analyze.")
# Sample data
with st.expander("Sample Data Format"):
st.write("""
Upload a CSV file with the following columns:
- student_response: Text responses from students
- misconception_label: Category of mathematical misconception (optional)
Example misconception categories:
- number_properties
- decimal_operations
- fraction_operations
- fraction_comparison
- decimal_comparison
""")
sample_data = pd.DataFrame({
'student_response': [
"I think negative numbers can't have square roots",
"When multiplying decimals, more decimal places means bigger number",
"Adding fractions means adding numerators and denominators separately",
"The larger the denominator, the larger the fraction",
"0.1 is greater than 0.9 because 1 is less than 9"
],
'misconception_label': [
'number_properties',
'decimal_operations',
'fraction_operations',
'fraction_comparison',
'decimal_comparison'
]
})
st.dataframe(sample_data)