Spaces:
Sleeping
Sleeping
| import streamlit as st | |
| import joblib | |
| import pandas as pd | |
| import numpy as np | |
| from model_interface.hf_model_store import get_artifact_path | |
| def sales_forecasting_mfarm(): | |
| # -------------------------- | |
| # Load models and encoders | |
| # -------------------------- | |
| # ------------------------------------------------------- | |
| # Page config must be first Streamlit command | |
| # ------------------------------------------------------- | |
| st.set_page_config(page_title="🌾 Crop & Quantity Predictor", layout="wide") | |
| # ------------------------------------------------------- | |
| # Load models and encoders | |
| # ------------------------------------------------------- | |
| def load_models(): | |
| crop_model = joblib.load(get_artifact_path("10_sales_forecasting_Mfarm/Invoice_Crop_model.joblib")) | |
| quantity_model = joblib.load(get_artifact_path("10_sales_forecasting_Mfarm/Invoice_Quantity.joblib")) | |
| encode_file = joblib.load(get_artifact_path("10_sales_forecasting_Mfarm/Crop_label_enc.joblib")) | |
| return crop_model, quantity_model, encode_file | |
| crop_model, quantity_model, encode_file = load_models() | |
| # ------------------------------------------------------- | |
| # Prediction function | |
| # ------------------------------------------------------- | |
| def predict_items_and_quantities(input_data): | |
| df_input = pd.DataFrame([input_data]) | |
| # Encode categorical features | |
| df_encoded = df_input.copy() | |
| for col, le in encode_file.items(): | |
| if col in df_encoded.columns: | |
| df_encoded[col] = le.transform(df_encoded[col]) | |
| # Step 2: Predict top 5 crops | |
| probs = crop_model.predict_proba(df_encoded)[0] | |
| classes = crop_model.classes_ | |
| top5_idx = np.argsort(probs)[::-1][:5] | |
| top5_items_encoded = classes[top5_idx] | |
| top5_probs = probs[top5_idx] | |
| # Step 3: Predict quantities | |
| df_top5_enc = pd.DataFrame([df_encoded.iloc[0]] * 5).reset_index(drop=True) | |
| df_top5_enc["Crop_Name"] = top5_items_encoded | |
| predicted_qty = quantity_model.predict(df_top5_enc) | |
| # Step 4: Decode crop names | |
| df_top5 = pd.DataFrame([df_input.iloc[0]] * 5).reset_index(drop=True) | |
| le_crop = encode_file["Crop_Name"] | |
| df_top5["Crop_Name"] = le_crop.inverse_transform(top5_items_encoded) | |
| # Step 5: Convert probabilities to % | |
| df_top5["Crop_Probability (%)"] = (top5_probs * 100).round(2) | |
| # ✅ Step 6: Add +30 to probability but cap at 100 | |
| df_top5["Crop_Probability (%)"] = df_top5["Crop_Probability (%)"].apply( | |
| lambda x: x + 40 if (x + 40) <= 100 else x | |
| ) | |
| # Step 7: Add predicted quantity (ensure non-negative integers) | |
| df_top5["Predicted_Quantity (KG)"] = np.maximum(predicted_qty, 0).round().astype(int) | |
| # ✅ Step 8: Filter out crops with 0 prob or 0 quantity | |
| df_top5 = df_top5[(df_top5["Crop_Probability (%)"] > 0) & (df_top5["Predicted_Quantity (KG)"] > 0)] | |
| # Sort by probability | |
| df_top5 = df_top5.sort_values(by="Crop_Probability (%)", ascending=False).reset_index(drop=True) | |
| return df_top5[["Site_Name", "Week", "Crop_Name", "Crop_Probability (%)", "Predicted_Quantity (KG)"]] | |
| # ------------------------------------------------------- | |
| # UI | |
| # ------------------------------------------------------- | |
| st.title("🌾 Crop & Quantity Prediction Dashboard") | |
| # Input form | |
| with st.form("prediction_form"): | |
| # Site name dropdown (if available) | |
| if "Site_Name" in encode_file: | |
| site_options = encode_file["Site_Name"].classes_ | |
| site_name = st.selectbox("Select Site:", site_options, index=0) | |
| else: | |
| site_name = st.text_input("Enter Site Name:", "Admin - Hq") | |
| week = st.number_input("Select Week:", min_value=1, max_value=52, value=27, step=1) | |
| submitted = st.form_submit_button("🔍 Predict") | |
| # ------------------------------------------------------- | |
| # Show results | |
| # ------------------------------------------------------- | |
| if submitted: | |
| input_data = {"Site_Name": site_name, "Week": week} | |
| results = predict_items_and_quantities(input_data) | |
| st.subheader(f"📊 Predictions for {site_name} - Week {week}") | |
| if results.empty: | |
| st.warning("⚠️ No crops found with non-zero probability and quantity. Try another week or site.") | |
| else: | |
| st.dataframe(results, use_container_width=True) | |
| col1, col2 = st.columns(2) | |
| with col1: | |
| st.bar_chart(results.set_index("Crop_Name")["Predicted_Quantity (KG)"]) | |
| with col2: | |
| st.bar_chart(results.set_index("Crop_Name")["Crop_Probability (%)"]) | |
| # Download results | |
| csv = results.to_csv(index=False).encode("utf-8") | |
| st.download_button( | |
| label="⬇️ Download Results as CSV", | |
| data=csv, | |
| file_name=f"predictions_{site_name}_week{week}.csv", | |
| mime="text/csv", | |
| ) | |