Quantifying Financial Emotion
Through Transaction Behavior

Parameter System IRS - PRS - Resilience: Turn raw data into comprehensible financial insights. Instead of just being an "Accountant", we build a personalized "Mental Coach".

The "Why"

Why does Emotion have to be Quantified in a financial app?

🚨 Current Problems

Recent financial apps only concentrate on reporting "Amount Spent" (What), yet skip "Spending Feeling" (Why/How).

The gap between unconcious "Paying" and the following "Regret" is the biggest cause of personal financial disorder.

🛡️ Solution: 3-Layer Protection

  • ⚡
    IRS (Immediate Regret Score): Immediate warning when the transaction is made.
  • 📅
    PRS (Periodic Regret Score): Reflect habits and standard deviation on a monthly basis.
  • 🌱
    Resilience Level: Measure long-term health and the ability to recover from shocks.

1. IRS - Immediate Regret Score

Measure regret probability (0 → 1) right after a single transaction, provide triggers (>0.65) ❓Why 0.65?
Our research points out that the proper threshold is 0.7; however, to cover all edge cases, we apply the "Weather Forcast" style, which means the output is "positive" for neutral cases. for LLM to intervene.

After a period of researching, we found out some factors that affect the IRS score:

Beyond Budget
How much is spent beyond planned budget.
Spent Amount
The ratio of the amount spent to the current income.
Goal Conflict
Conflict with current saving goals.
Transaction Time
Late night transaction (11 PM - 5 AM).

🧠 Interaction Amplifiers

Psychology-related value is non-linear. When one parameter cross the threshold, the IRS forms an exponential function.

Therefore, we need Interaction Amplifiers to standardize how fast IRS increases in one single case.

🎛️ IRS Simulator 0.12

Move the slide bar to see how our model interacts with transaction behavior.

0%
Low
Day (0)

Turn on to emulate buying at 2 AM.

2. PRS - Periodic Regret Score

Evaluate accumulated negative feeling and the deviation in spending in a period (usually monthly).

Factors that affect PRS:

Impulsive Spending
The rate of amount spent on unnecessary categories.
Goal Conflict
Conflict with current saving goals.
Financial Pressure
Real pressure: debt, low income, ect.

🧠 Tempporal Discounting

Users tend to prioritize immediate prizes (impulsive buying) instead of future plans (saving). PRS reveals this trend using long-term charts.

Regret vs. Savings Trend (6 Months)

When PRS climbs, real Savings Rate tends to go down considerably.

3. Resilience

Evaluate financial "immune system". Resilience doesn't mean never wasting, it means the ability to rebalance the money flow and and be disciplined again after a financial shock.

Factors to classify users (Struggling or Disciplined) using Centering Bias δ = 5.5.

Recovery Speed
How fast the user restores safe balance.
Spending Entropy
The level of disorder in transaction behavior.
Income Volatility
Stability and shock tolerance.
Goal Consistency
Monthly goal saving frequency.

🧠 Financial Self-Efficacy

High Resilience Score (>0.9) proves that one user has enough faith and ability to control financial situation, encouraging LLM to propose investment with higher risks.

Future Scope

🚧 Limitations

  • 🏷️
    Category Tagging: Completely rely on mannually labeling categories. If transactions are mislabeled, IRS/PRS will be interfered.
  • ⏳
    Data Cold-Start: Need at least 3 months of data for PRS and Resilience Level to reach best accuracy.

🚀 Development

  • 🤖
    Hyper-Personalization: System automatically learns and adjusts weights (α, β) based on real users feedbacks, combined with personality classifying by Machine Learning.
  • 🤝
    Interconnection: We look forward to collaboration from multiple parties to obtain more data and improve accuracy of our model.