You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Pickle files are version-sensitive. Loading with a different xgboost version will crash or give wrong predictions.

Usage (direct)

import pickle
from huggingface_hub import hf_hub_download

model_path = hf_hub_download(
    repo_id="hossamaladdin/feasbillty_check_travel",
    filename="feasibility_model.pkl"
)
encoder_path = hf_hub_download(
    repo_id="hossamaladdin/feasbillty_check_travel",
    filename="label_encoder.pkl"
)

with open(model_path, "rb") as f:
    model = pickle.load(f)

with open(encoder_path, "rb") as f:
    le = pickle.load(f)

features = [[5, 2, 1000.0, 180.0, 1800.0, 0.556, 3, 1, 0]]
pred = model.predict(features)
print(le.inverse_transform(pred))

Recommended Usage (with bridge layer)

Do not call the model directly with raw NER output. Use entity_handler.py to convert NER entities to numeric features first.

from backend.entity_handler import entity_to_features, get_diagnostic_tags, explain_feasibility

entities = {
    "LOCATION":    "Paris",
    "DURATION":    "5 days",
    "BUDGET":      "1000 dollars",
    "GROUP_SIZE":  "2 people",
    "TRAVEL_TYPE": "romantic",
    "DATE":        "in July"
}

features = entity_to_features(entities)
pred = model.predict([list(features.values())])
label = le.inverse_transform(pred)[0]
tags = get_diagnostic_tags(features)
explanation = explain_feasibility(label, tags)

Performance

Test Accuracy : 82.47% Weighted F1 : 81.90% Macro F1 : 79.17%

Class	Precision	Recall	F1	Support
feasible	0.76	0.97	0.85	293
partial	0.96	0.76	0.85	197
infeasible	0.85	0.56	0.67	109

Known limitation: infeasible recall = 56%. Model tends to classify borderline infeasible trips as partial. This is an acceptable failure mode — user still receives a warning.

Downloads last month: -; Downloads are not tracked for this model. How to track