Case File
Personality-Based Place Recommendation & Routing System
Survey data linked personality dimensions to place labels; classifiers were compared on the same feature matrix.
- Domain
- Recommendation Systems
- Focus
- Model comparison, data analysis
- Stack
- Python, Scikit-learn, Pandas
- Status
- Completed
- Year
- 2025
Case cover
Feature importance (RIASEC and demographic predictors)
Problem
Preferences are mostly implicit; inference had to come from surveys and labels rather than direct scores. Generalization of competing classifiers was unknown upfront.
Approach
Fixed the feature matrix and used a reproducible train/hold-out split. Metrics: accuracy, F1, confusion matrix; weighted F1 under class imbalance.
Technical Structure
Data preparation
Pulled personality and place-preference fields from surveys; dropped incomplete or inconsistent rows.
Feature logic
Big-Five-like dimensions numerically encoded; venue types one-hot or ordinal encoded.
Model comparison
Logistic Regression, Random Forest, Gradient Boosting on the same split; small grid search only.
Evaluation metrics
Accuracy, precision, recall, macro/micro F1; default decision threshold 0.5.
Outcome
Selected the top F1 model; feature set and encoding scheme documented for reruns.
Model comparison
| Metric | Scenario | Decision Trees | Random Forest | Boosting Methods | |||
|---|---|---|---|---|---|---|---|
| Decision Trees | Multi Output Classifier | Multi Output Regressor | Classifier Chain | XGBoost | LightGBM | ||
| Mikro F-1 | Standart Model | 0.6270 | 0.6769 | 0.6734 | 0.6410 | 0.6570 | 0.6772 |
| Demography | 0.6599 | 0.7247 | 0.7229 | 0.6938 | 0.7086 | 0.6807 | |
| Hyperparameter | - | 0.7311 | 0.7288 | - | 0.7149 | 0.7287 | |
| Makro F-1 | Standart Model | 0.5804 | 0.5823 | 0.6227 | 0.5772 | 0.6017 | 0.6285 |
| Demography | 0.6205 | 0.6492 | 0.6564 | 0.6350 | 0.6545 | 0.6297 | |
| Hyperparameter | - | 0.6529 | 0.6495 | - | 0.6543 | 0.6792 | |
| Hamming Loss | Standart Model | 0.2812 | 0.2211 | 0.2326 | 0.2402 | 0.2454 | 0.2297 |
| Demography | 0.2538 | 0.1895 | 0.1968 | 0.2046 | 0.2092 | 0.2237 | |
| Hyperparameter | - | 0.1861 | 0.1887 | - | 0.1994 | 0.1956 | |