Dashboard

PocketSmart AI Engine is a rule-based AI: it understands questions by intent matching and uses statistics for forecasts and unusual-spend detection. It runs fully offline and doesn't call an external AI model.

• Budget: income is split by your category percentages (default 50/30/20 rule).
• Categorisation: keywords in the description (Swiggy → Food, Uber → Travel, EB → Bills).
• Overspending: over budget when spent > budget; near limit at 80%; likely to overspend when the month-end projection is above budget.
• Month-end projection: spent so far + remaining share of the month × last month's amount. Fixed bills use last month's amount.
• Forecast: weighted average of this month (projected) and the two previous months with weights 50%, 30% and 20%.
• Savings goal: needed per month = (target − saved) ÷ months left.
• Recommendations: suggested cut = overspend, capped at 25% of the category's spending.
• Categorisation confidence: brand name matched 95%, general keyword 80%, no match 40%.
• Unusual spend: a transaction more than 2× the category's average transaction over the last 3 months (rent excluded).
• Forecast confidence: 100 − variation of the last 3 months' totals (standard deviation ÷ average), kept between 60% and 98%.
• Financial health score: savings rate 40 + budget discipline 30 + goal progress 20 + spending trend 10.