Forecasting a week of retail sales

Version: 2026-09-21

The prepared Rossmann table contains 942 daily records from 1 January 2013 to 31 July 2015. It is an aggregate retail series: there is one row per date, no store identifier, and the open and customers fields contain daily counts. The model therefore forecasts the supplied aggregate sales series, rather than a specific individual store. Sales retain the source file units; the prepared CSV does not label a currency.

This is an internal historical benchmark. The chronological windows do not cross subset boundaries, but the native preprocessing statistics use the saved used-row set, including later periods. A prospective evaluation must fit preprocessing within training and roll the forecast origin forward. Overlapping test windows are dependent. Results do not establish transfer to a different store, future period or promotional policy.

Open the .nd project in Neural Designer. Its embedded data, parameters, saved sample roles and task report are preserved. If the original CSV path is unavailable, select the CSV included here. The source CSV may include unused columns; the embedded project defines the exact modelling schema.

The charts/ directory contains standalone HTML exported with the application chart builder. The schema and test-metrics files identify the exact inputs, outputs, evidence and source hash.

Dataset source: https://www.kaggle.com/c/rossmann-store-sales/data
Retain the source attribution and its applicable dataset terms.

The verified inference.js provides the exported model arithmetic with a seven-horizon return adapter.

Correction: the supplied project predicted school_holiday. This copy predicts sales, using native Neural Engine import, training, neuron selection and testing with seed 20260921. The chronological windows remain within their source segments.

Forecast contract: 70 inputs = 7 days x 10 fields, day-major from oldest to newest. See input-schema.json. Outputs are seven daily sales values.

The native HTML computed all seven horizons but returned only the last. The included inference.js changes only that return list and adds input validation; all generated model arithmetic is retained. All 176 x 7 outputs were independently checked against the saved recurrent parameter buffer and native report. The incomplete original HTML is not included as a deployment tool.

Source: Rossmann Store Sales, https://www.kaggle.com/c/rossmann-store-sales/data . The supplied CSV is a prepared daily aggregate. Sales retain source units. These forecasts do not estimate causal promotional uplift.

Temporal means update: the saved Calculate temporal means task from the user project has been appended unchanged. It summarizes the same 942 source records. The model configuration, parameters, data, earlier reports and inference.js are unchanged. The three sales charts are native Neural Viewer HTML exports.
