Create pages/5 Model Building.py
Browse files- pages/5 Model Building.py +52 -0
pages/5 Model Building.py
ADDED
|
@@ -0,0 +1,52 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import streamlit as st
|
| 2 |
+
|
| 3 |
+
# βοΈ Model Building & Evaluation
|
| 4 |
+
st.markdown("""
|
| 5 |
+
## βοΈ Model Building & Evaluation
|
| 6 |
+
|
| 7 |
+
### ποΈ Model Building:
|
| 8 |
+
|
| 9 |
+
The classification model was built using the **KNN Classifier**, a robust ensemble learning method suitable for handling both numerical and categorical data.
|
| 10 |
+
|
| 11 |
+
**Model Pipeline:**
|
| 12 |
+
|
| 13 |
+
- **Preprocessing**:
|
| 14 |
+
- Applied **ColumnTransformer** to handle both categorical and numerical features.
|
| 15 |
+
- Used **StandardScaler** for scaling continuous variables.
|
| 16 |
+
- Encoded categorical columns using **OrdinalEncoder** (where appropriate).
|
| 17 |
+
|
| 18 |
+
- **Model**:
|
| 19 |
+
- Trained using **KNearestNeighbors Classifier** from `scikit-learn` with tuned hyperparameters.
|
| 20 |
+
|
| 21 |
+
---
|
| 22 |
+
|
| 23 |
+
### β
Model Training:
|
| 24 |
+
|
| 25 |
+
- The dataset was split into **training and testing sets** using an 80-20 split.
|
| 26 |
+
- The model was trained on the training set using the engineered features.
|
| 27 |
+
- Proper cross-validation techniques were used to avoid overfitting.
|
| 28 |
+
|
| 29 |
+
---
|
| 30 |
+
|
| 31 |
+
### π Model Evaluation:
|
| 32 |
+
|
| 33 |
+
**Evaluation Metrics Used:**
|
| 34 |
+
|
| 35 |
+
- **Accuracy** β Measures the overall correctness of the model.
|
| 36 |
+
- **Precision** β Measures how many predicted "High Risk" projects were actually high risk.
|
| 37 |
+
- **Recall** β Indicates how many actual high-risk projects were correctly identified.
|
| 38 |
+
- **F1 Score** β Harmonic mean of precision and recall for balanced performance.
|
| 39 |
+
- **Confusion Matrix** β Visual breakdown of correct and incorrect classifications.
|
| 40 |
+
|
| 41 |
+
---
|
| 42 |
+
|
| 43 |
+
The model demonstrated **strong classification performance** on the test dataset, achieving good scores across all metrics.
|
| 44 |
+
It was then saved as a `.pkl` file for deployment and integrated into the Streamlit app for real-time predictions.
|
| 45 |
+
""")
|
| 46 |
+
|
| 47 |
+
|
| 48 |
+
if st.button('Go to Deployment'):
|
| 49 |
+
st.switch_page(r"pages\6.Deployment.py")
|
| 50 |
+
|
| 51 |
+
if st.button('Back'):
|
| 52 |
+
st.switch_page(r"pages\4.Feature Engineering.py")
|