Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Integrating AI and Ensemble Forecasting: Explainable Materials Planning with Scorecards and Trend Insights for a Large-Scale Manufacturer

Saravanan Venkatachalam

TL;DR

This work tackles after-sales spare-parts demand forecasting for a large automotive manufacturer with long-tailed, intermittent demand across 90+ countries and ~6,000 parts. It proposes a revenue- and cluster-aware ensemble that blends statistical, ML, and DL forecasters with a private LLM-based analytics layer to produce role-specific narratives. Key contributions include a horizon-aware ensemble with nonnegative weights learned by rolling-origin validation, a Pareto-aware segmentation strategy, calibrated prediction intervals, and a context-engineering framework that translates forecasts into $MAPE$/$WMAPE$-weighted scorecards and trend insights, with calibrated predictive intervals. The architecture ensures governance, reproducibility, and a closed loop from forecasts to inventory decisions across multiple regions, scalable to new regimes and markets.

Abstract

This paper presents a practical architecture for after-sales demand forecasting and monitoring that unifies a revenue- and cluster-aware ensemble of statistical, machine-learning, and deep-learning models with a role-driven analytics layer for scorecards and trend diagnostics. The framework ingests exogenous signals (installed base, pricing, macro indicators, life cycle, seasonality) and treats COVID-19 as a distinct regime, producing country-part forecasts with calibrated intervals. A Pareto-aware segmentation forecasts high-revenue items individually and pools the long tail via clusters, while horizon-aware ensembling aligns weights with business-relevant losses (e.g., WMAPE). Beyond forecasts, a performance scorecard delivers decision-focused insights: accuracy within tolerance thresholds by revenue share and count, bias decomposition (over- vs under-forecast), geographic and product-family hotspots, and ranked root causes tied to high-impact part-country pairs. A trend module tracks trajectories of MAPE/WMAPE and bias across recent months, flags entities that are improving or deteriorating, detects change points aligned with known regimes, and attributes movements to lifecycle and seasonal factors. LLMs are embedded in the analytics layer to generate role-aware narratives and enforce reporting contracts. They standardize business definitions, automate quality checks and reconciliations, and translate quantitative results into concise, explainable summaries for planners and executives. The system exposes a reproducible workflow -- request specification, model execution, database-backed artifacts, and AI-generated narratives -- so planners can move from "How accurate are we now?" to "Where is accuracy heading and which levers should we pull?", closing the loop between forecasting, monitoring, and inventory decisions across more than 90 countries and about 6,000 parts.

Integrating AI and Ensemble Forecasting: Explainable Materials Planning with Scorecards and Trend Insights for a Large-Scale Manufacturer

TL;DR

This work tackles after-sales spare-parts demand forecasting for a large automotive manufacturer with long-tailed, intermittent demand across 90+ countries and ~6,000 parts. It proposes a revenue- and cluster-aware ensemble that blends statistical, ML, and DL forecasters with a private LLM-based analytics layer to produce role-specific narratives. Key contributions include a horizon-aware ensemble with nonnegative weights learned by rolling-origin validation, a Pareto-aware segmentation strategy, calibrated prediction intervals, and a context-engineering framework that translates forecasts into /-weighted scorecards and trend insights, with calibrated predictive intervals. The architecture ensures governance, reproducibility, and a closed loop from forecasts to inventory decisions across multiple regions, scalable to new regimes and markets.

Abstract

This paper presents a practical architecture for after-sales demand forecasting and monitoring that unifies a revenue- and cluster-aware ensemble of statistical, machine-learning, and deep-learning models with a role-driven analytics layer for scorecards and trend diagnostics. The framework ingests exogenous signals (installed base, pricing, macro indicators, life cycle, seasonality) and treats COVID-19 as a distinct regime, producing country-part forecasts with calibrated intervals. A Pareto-aware segmentation forecasts high-revenue items individually and pools the long tail via clusters, while horizon-aware ensembling aligns weights with business-relevant losses (e.g., WMAPE). Beyond forecasts, a performance scorecard delivers decision-focused insights: accuracy within tolerance thresholds by revenue share and count, bias decomposition (over- vs under-forecast), geographic and product-family hotspots, and ranked root causes tied to high-impact part-country pairs. A trend module tracks trajectories of MAPE/WMAPE and bias across recent months, flags entities that are improving or deteriorating, detects change points aligned with known regimes, and attributes movements to lifecycle and seasonal factors. LLMs are embedded in the analytics layer to generate role-aware narratives and enforce reporting contracts. They standardize business definitions, automate quality checks and reconciliations, and translate quantitative results into concise, explainable summaries for planners and executives. The system exposes a reproducible workflow -- request specification, model execution, database-backed artifacts, and AI-generated narratives -- so planners can move from "How accurate are we now?" to "Where is accuracy heading and which levers should we pull?", closing the loop between forecasting, monitoring, and inventory decisions across more than 90 countries and about 6,000 parts.

Paper Structure

This paper contains 14 sections, 4 figures.

Table of Contents

  1. Introduction
  2. Background and Motivation
  3. Forecasting Methods
  4. Large Language Models
  5. Description
  6. Technical Architecture
  7. Implementation
  8. Dashboard
  9. Context Engineering Framework
  10. Performance Analysis using AI
  11. Trend Analysis using AI
  12. Trend Analysis using AI - Overall
  13. Trend Analysis using AI - Monthwise
  14. Conclusion and Future Directions

Figures (4)

  • Figure 1: The diagram shows two AI agents—one for business insights and the other for algorithm insights—the agents are positioned between the users and the forecasting algorithm. The forecasting engine (an ensemble of statistical, ML, and deep-learning methods) ingests history and exogenous variables, and produces forecasts and predictive intervals. The business user AI agent provides perspectives on business actions (opportunities/risks, performance bins, revenue mix, and next steps), while the other AI agent offers insights on technical diagnostics (error metrics, distributions, drivers, and improvement ideas).
  • Figure 2: System architecture for descriptive business insights and forecasting with AI agents, optimization engines, and dynamic result generation.
  • Figure 3: Forecast Performance Dashboard interface displaying real-time metrics on various metrics, insights, and recommendations for next steps. The dashboard enables users to get insights on performance and trend for materials planning related to forecasting and safety stocks.
  • Figure 4: Context engineering architecture integrating LLMs, REST APIs, and data systems to deliver interactive, explainable, and role-specific decision support in seven structured steps.