Accounting Graph Transformer for Short-History Multi-KPI Forecasting in Small Businesses

2026-08-07

AI Summary

This study addresses the challenge of jointly forecasting 13 key performance indicators (KPIs) across financial statements over a 12-month horizon for small businesses with only 12–24 months of accounting history. To this end, the authors propose the Accounting Graph Transformer (AGT), the first model to embed structured accounting relationship graphs into a Transformer architecture. AGT employs a typed graph attention mechanism to propagate information over fixed accounting relationships and integrates target-specific contextual cues with a gated representation of the most recent three months’ trajectories, enabling unified, company-agnostic multi-KPI forecasting. Evaluated on 1,060 unseen firms, AGT achieves a macro MAE of 0.6990, significantly outperforming baselines such as LightGBM across all 13 KPIs, and demonstrates strong generalization on an additional hold-out set of 7,094 companies.

Abstract

Small businesses often have only 12-24 months of accounting history, yet planning and risk workflows require coordinated forecasts across financial statements. We study joint 12-month forecasting of 13 income-statement, balance-sheet, cash-flow, and working-capital key performance indicators (KPIs) from 71 monthly ledger series. We introduce the Accounting Graph Transformer (AGT), which represents each ledger series as a masked token, exchanges information through typed attention on a fixed accounting-relation graph, pools target-specific context, and fuses it with a gated three-month recency path. Across 11,993 forecast origins from 1,060 unseen companies, AGT achieves sample-weighted KPI-macro mean absolute error (MAE) $0.6990 \pm 0.0013$ over three independent seeds, compared with $0.7378 \pm 0.0014$ for the strongest baseline, LightGBM. At the pre-specified seed 42, a paired company-clustered bootstrap gives a LightGBM-minus-AGT difference of 0.0395 with 95% confidence interval (CI) $[0.0350,0.0439]$. AGT is best on all 13 KPIs against LightGBM, TimeMixer, and SOFTS in the matched seed-42 comparison, while final-architecture ablations show that relational attention, accounting topology, and the recency path each improve validation and test accuracy. On 7,094 additional unseen companies with origins sampled from January-May 2025, AGT obtains 0.7548 MAE versus 0.7694 for SOFTS. A single 5.3M-parameter model produces 156 aligned forecasts without company-specific fitting, providing one forecasting layer for integrated planning, liquidity, and working-capital analysis.

Problem

• short-history forecasting
• multi-KPI forecasting
• small businesses
• financial statement forecasting
• accounting time series

Innovation

• Accounting Graph Transformer
• short-history forecasting
• multi-KPI joint prediction
• typed attention
• accounting relation graph