Paper: arXiv 2306.07305

Authors: Shaun D’Souza, Dheeraj Shah, Amareshwar Allati, Parikshit Soni

Abstract

Retail sales and price projections are typically based on time series forecasting. For some product categories, the accuracy of demand forecasts achieved is low, negatively impacting inventory, transport, and replenishment planning. This paper presents our findings based on a proactive pilot exercise to explore ways to help retailers to improve forecast accuracy for such product categories. We evaluated opportunities for algorithmic interventions to improve forecast accuracy based on a sample product category, Knitwear. The Knitwear product category has a current demand forecast accuracy from non-AI models in the range of 60%. We explored how to improve the forecast accuracy using a rack approach. To generate forecasts, our decision model dynamically selects the best algorithm from an algorithm rack based on performance for a given state and context. Outcomes from our AI/ML forecasting model built using advanced feature engineering show an increase in the accuracy of demand forecast for Knitwear product category by 20%, taking the overall accuracy to 80%. Because our rack comprises algorithms that cater to a range of customer data sets, the forecasting model can be easily tailored for specific customer contexts.

Complexity vs Empirical Score

  • Math Complexity: 3.5/10
  • Empirical Rigor: 6.5/10
  • Quadrant: Street Traders — empirical and applied, lower math complexity

Why this score: The paper relies on standard ML techniques (XGBoost, LSTM) with minimal advanced math, but demonstrates strong empirical implementation using a real retail dataset, feature engineering, and clear accuracy metrics (RMSE, R^2).

Research Flowchart

  flowchart TD
  A["Research Goal: Improve demand forecast accuracy for low-accuracy categories like Knitwear (60%)"] --> B["Methodology: Algorithmic Rack Approach"]
  B --> C["Data Input: Historical sales & price data for Knitwear"]
  C --> D["Computational Process: Dynamic algorithm selection based on state/context"]
  D --> E["Advanced Feature Engineering & AI/ML Model Training"]
  E --> F["Outcomes: Forecast accuracy improved by 20% to 80%"]
  F --> G["Benefit: Model is adaptable to diverse customer datasets"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes an algorithmic rack of custom AI/ML models with a decision model that dynamically selects the best‑fit forecasting algorithm for a given state and context.

Source sentence
We created an algorithmic rack comprising multiple custom AI/ML based models for demand prediction and designed a decision model for dynamic selection of best-fit algorithm for given state and context.

Assumes 1

  • The past sales data can guide future sales forecasts. model
    Source sentence
    It assumes that the past is a guidance into the future sales forecasts and can forecast any major demand shifts in the inventories arising out of seasonality’s such as upcoming festivals.

Data

  • Knitwear sales data alternative unknown
    Source sentence
    Table 1: Sales data for Knitwear
  • Knitwear demand dataset (≈3 million points) alternative unknown
    Source sentence
    It highlights the actual vs. predicted trend for our dataset with ∼3 million data points.

Costs

Not applicable The paper reports no trading or execution result.

Finds 3

  • The AI/ML forecasting model improves Knitwear demand forecast accuracy by 20 percentage points, raising overall accuracy to 80 %. predictive-accuracy
    Source sentence
    Outcomes from our AI/ML forecasting model built using advanced feature engineering show an increase in the accuracy of demand forecast for Knitwear product category by 20%, taking the overall accuracy to 80%.
  • XGBoost Regressor yields the smallest RMSE and high R² on train and test sets, making it the most suitable regression model. performance
    Source sentence
    The data shows that the XGBoost Regressor produces the smallest RMSE and high R2 in the train and test sets, making it the most suitable for our regression modeling.
  • The decision‑tree based dynamic model selector can automatically choose the forecasting algorithm with the least error for each context. predictive-accuracy
    Source sentence
    ... model selector ←sklearn.tree.DecisionTreeClassifier().fit(X, y)

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 16 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.