About Prompt
- Prompt Type – Dynamic
- Prompt Platform – ChatGPT, Grok, Deepseek, Gemini, Copilot, Midjourney, Meta AI and more
- Niche – Behavioral Credit Scoring
- Language – English
- Category – Credit Risk Analytics
- Prompt Title – AI Prompt for Calculating Credit Risk Score Based on Behavioral Data
Prompt Details
**Goal:** Develop a robust credit risk scoring model utilizing behavioral data to enhance traditional credit scoring methods for improved risk assessment.
**Prompt Type:** Dynamic
**Target Platform:** All AI Platforms (adaptable)
**Input Data Format:** Assume access to a structured dataset containing behavioral and traditional credit data. Specify the desired format (e.g., CSV, JSON) when using the prompt with a specific platform. The dataset should include, but isn’t limited to:
* **Traditional Credit Data (if available):**
* Age
* Income
* Employment history
* Existing credit lines and balances
* Credit utilization
* Payment history (number of late payments, delinquencies)
* Credit inquiries
* **Behavioral Data:**
* Transaction history (frequency, amount, type, merchant category)
* Online browsing behavior (e.g., websites visited, products viewed, time spent)
* Social media activity (if ethically sourced and permissible) (e.g., connections, posts, engagement)
* Mobile app usage (e.g., time spent on financial apps, budgeting behavior)
* Geolocation data (if ethically sourced and permissible)
* Device usage patterns (e.g., type of device, operating system)
* Customer service interactions (e.g., frequency, sentiment)
**Output Requirements:**
1. **Credit Risk Score:** Generate a numerical credit risk score for each individual in the dataset, ranging from [lowest score] to [highest score], where a higher score indicates lower risk. Define the score range and its interpretation based on the specific use case (e.g., 300-850, similar to FICO scores, or a simpler 1-10 scale).
2. **Feature Importance:** Provide a ranked list of the most influential features contributing to the credit risk score. This helps understand the model’s decision-making process and identify key behavioral indicators of creditworthiness.
3. **Model Explainability (Optional but Highly Recommended):** If the platform supports it, generate explanations for individual credit risk scores. This could involve highlighting specific behavioral patterns or data points that contributed significantly to the assigned score, enhancing transparency and trust.
4. **Model Performance Metrics:** Evaluate the model’s performance using appropriate metrics like AUC-ROC, precision, recall, F1-score, and KS statistic. Include a confusion matrix if applicable.
5. **Model Robustness Evaluation (Optional):** Assess the model’s robustness to potential biases in the data (e.g., demographic biases). Suggest mitigation strategies if any biases are detected.
**Prompt Structure (Adaptable):**
“`
Task: Develop a credit risk scoring model using the provided dataset.
Dataset: [Provide dataset path/name or directly input the data]
Data Format: [Specify data format – e.g., CSV, JSON]
Score Range: [Define the desired score range – e.g., 300-850, 1-10]
Output:
1. Credit Risk Score: Calculate a credit risk score for each individual in the dataset within the specified range.
2. Feature Importance: Rank the features by their contribution to the score.
3. Model Explainability (If possible): Provide explanations for individual scores.
4. Model Performance Metrics: Evaluate model performance using AUC-ROC, precision, recall, F1-score, and KS statistic. Include a confusion matrix.
5. Model Robustness Evaluation (Optional): Assess for biases and suggest mitigation strategies.
Model Type (Optional): [Specify a preferred model type if desired – e.g., gradient boosting, neural network]. Allow the AI to choose the best model if no specific type is provided.
Constraints: [Specify any constraints – e.g., computational resources, explainability requirements].
“`
**Example for a specific platform (Python with scikit-learn):**
“`python
# Import necessary libraries …
# Load the dataset …
# Define the prompt (as a string or dictionary depending on the platform)…
# Execute the prompt using the chosen platform’s API …
# Process and interpret the results …
“`
**Important Notes:**
* Adapt this prompt to the specific AI platform you are using.
* Clearly define the score range and its interpretation.
* Prioritize ethical considerations regarding data privacy and potential biases.
* Experiment with different model parameters and feature engineering techniques to optimize performance.
* Regularly retrain and evaluate the model with new data to maintain accuracy and adapt to evolving behavioral patterns.