Supplier Intelligence & Vendor Management Portal — Final No-Limits Build

This is the final thesis-aligned localhost application for the 7005SCN supplier-performance project.

Preloaded research data

The application includes the complete processed UK Contracts Finder dataset supplied for 2021–2025:

124,781 unique suppliers
179,114 supplier-year evaluation records
2021: 33,987 supplier-year rows
2022: 35,540 supplier-year rows
2023: 40,108 supplier-year rows
2024: 40,428 supplier-year rows
2025: 29,051 supplier-year rows

Operational supplier-performance variables that are not observed in Contracts Finder remain labelled Synthetic in accordance with the research methodology.

Artificial upper limits removed

This build does not impose an application-level upper row cap on:

KPI CSV imports
purchase-order CSV imports
Contracts Finder .csv.tar.gz, .xlsx or .jsonl.gz imports
direct Contracts Finder 2021–2025 imports
machine-learning training CSVs
supplier records
evaluation records
purchase-order records
audit records
exported CSV records

Large tables are paginated in the interface only for browser performance. Pagination does not remove data and TOPSIS ranking operates on the complete active supplier set.

Included functionality

Supplier register and searchable supplier selection
Purchase-order/vendor-management records
Seven supplier KPIs
Historical TOPSIS ranking
Predictive TOPSIS ranking
Next-order risk prediction
Future performance-score regression
Pretrained model from the prepared Contracts Finder training-pair dataset
Role-based access
Audit trail
Full CSV export
Manual and official dataset import

Start on Windows

Run:

start_windows.bat

The launcher now waits until the local Flask server is ready before opening the browser.

On first launch, create the Owner account. The preloaded research database contains no preset user credentials.

Model Lab: raw Contracts Finder vs ML training data

The official releases_YYYY.csv files are raw procurement-release tables. They do not contain the derived supplier-performance inputs (price, quality, delivery, defects, response, returns, fulfilment) or future model targets (next_risk, next_performance) required for supervised ML training.

For this dissertation, use Train preloaded Contracts Finder 2021–2025 in Model Lab. It trains the models from the prepared supplier-year training table bundled with the application. Raw Contracts Finder archives remain available for procurement-data ingestion through Data Import.

Preloaded purchase orders

This final build includes 60 synthetic purchase-order records across 30 suppliers so the Purchase Orders module and purchase-order-derived KPI logic can be analysed immediately. These records are explicitly labelled Synthetic and are intended for prototype/thesis analysis rather than representation as observed UK procurement outcomes.

7006SCN2526MAYJUL/VENDOR_MANAGEMENT

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Languages