agricultural-prediction-on-viya

Overview

This repository provides a comprehensive SAS Viya pipeline for agricultural parcel analysis, crop prediction, and compliance analytics. The solution automates data cleaning, feature engineering, binary and multi-class crop modeling, and delivers ready-to-use tables for advanced reporting and SAS Visual Analytics dashboards.

Pipeline Execution Order

1. 01_data_cleaning_and_categorization.sas
   Cleans raw parcel data, handles Turkish characters, and performs initial NVDI and compliance checks. Outputs: final_promoted

2. 02_urunler_analysis_and_filtering.sas
   Analyzes and cleans product declarations, categorizes fallow land, and prepares product frequency tables. Outputs: parcel_analysis_promoted, product_frequency_promoted, analysis_summary_promoted

3. 03_product_categorization.sas
   Categorizes parcels by crop type and economic value, creates model-ready features. Outputs: parcel_model_ready_promoted, category_model_summary_promoted

4. 04_unregistered_parcels_analysis.sas
   Isolates CKS-unregistered parcels for scoring. Output: unregistered_parcels_promoted

5. 05_unregistered_product_categorization.sas
   Adds model-compatible features to unregistered parcels. Outputs: unreg_model_ready, unreg_summary

6. Model Studio (Binary Classification)
   Trains a model to detect production on unregistered parcels.

   • Training: parcel_model_ready_promoted
   • Scoring: unreg_model_ready
   • Output: unreg_table_scored

7. analysis/scored_table_analysis.sas
   Analyzes binary model results, risk categories, and prepares scoring outputs. Outputs: scoring_analysis_promoted, mahalle_analysis_promoted, risk_categories_promoted

8. isolate_44k.sas
   Isolates ~44K parcels predicted to have production and prepares crop prediction training data. Outputs: suspected_44k_promoted, crop_training_data_promoted

9. Model Studio (Crop Prediction)
   Trains a multi-class model to predict specific crop types.

   • Training: crop_training_data_promoted (target: crop_target)
   • Scoring: suspected_44k_promoted
   • Output: predicted_44k_products_final

10. analysis/crop_prediction_analysis.sas
    Analyzes crop prediction results, creates summary and neighborhood-level tables. Outputs: crop_analysis_promoted, category_summary_promoted, mahalle_crop_promoted

Visual Analytics-Ready Output Tables

• final_promoted: Cleaned, feature-rich master parcel table
• parcel_analysis_promoted: CKS-registered parcel details and product analysis
• category_model_summary_promoted: Crop category summary statistics
• scoring_analysis_promoted: Binary model scoring results for unregistered parcels
• mahalle_analysis_promoted: Neighborhood-level risk and production analysis
• risk_categories_promoted: Risk level summaries for dashboarding
• product_frequency_promoted: Product frequency and area analysis
• analysis_summary_promoted: General summary statistics
• parcel_model_ready_promoted: Model-ready CKS-registered parcels
• unreg_model_ready: Model-ready CKS-unregistered parcels
• unreg_summary: Summary for unregistered parcels
• crop_analysis_promoted: Parcel-level crop prediction results (specific crop names)
• category_summary_promoted: Crop frequency and area summaries (for VA frequency charts)
• mahalle_crop_promoted: Neighborhood-level dominant crop analysis

Using the Outputs in SAS Visual Analytics

All promoted tables are available in the casuser library and are ready for direct use in SAS Visual Analytics. Recommended visualizations include:

• Bar charts and pie charts for crop frequency (category_summary_promoted)
• Bubble plots for crop frequency vs. area vs. confidence
• List tables for detailed parcel or product analysis
• Geographic maps using mahalle and crop predictions