Ollama Expense Forecasting Tool
Load your CSV expense file and get a statistical forecast for the next 1–6 months using Prophet, anomaly detection with ML (IsolationForest), and an optional AI narrative about savings all processed locally: your data never leaves your private server.
Load your CSV file
Expected format: columns date, amount, category with header in the first row. Dates in format YYYY-MM-DD, amounts in euros.
Come utilizzare Expense Forecaster AI
Load your transaction CSV file
Drag and select a CSV file with date columns, amounts, and categories. You can download an example CSV to see the expected format.
Verify preview and spending by category
Check the first imported lines and average monthly spend by category before launching the forecast.
Configure Horizon and AI Narrative
Choose your forecast horizon (1-6 months) and decide if to include generated narrative from Ollama on-prem with savings suggestions.
Analyze forecast and anomalies
The tool returns monthly forecast with confidence interval (Prophet), anomaly transactions (Isolation Forest), and optional narrative.
Suggerimenti
- More months of historical data you provide, the more accurate your forecast will be: with less than 3 months of data, the model automatically switches to a simplified seasonal forecast.
- Categorize transactions consistently (recurring labels): helps both category aggregation and anomaly detection.
- Disable AI narrative if you only want the statistical forecast: speed up processing and don't require active TBA.
Domande frequenti
What format should the CSV file have?
The CSV file must have a header row with columns for date (YYYY-MM-DD format), amount (in euros) and category (free text). Use the "Download Example CSV" button for a pre-filled template.
How does statistical forecasting work?
The forecast uses the Prophet model (or seasonal-naive for short-term historical data), applied to your historical expense series, returning a monthly prediction with an 80% confidence interval.
How are anomalies detected?
The Isolation Forest algorithm analyzes your spending history profile and identifies statistically unusual transactions relative to your habits, assigning an anomaly score to each.
Are my financial data secure?
Yes. All processing occurs on-prem on a private Hetzner server with local Ollama: no data is shared with OpenAI, Anthropic or other third-party cloud services.
Can I trust AI narrative for financial decisions?
The narrative is generated by a local AI model for informational purposes and does not replace the advice of a professional. Always verify suggestions with a qualified financial advisor before acting.