Edit model card

Patient Readmission Prediction

Tranning

Github: prabinpanta0/Patient-Readmission-Prediction

Dataset

{
  "model_id": "prabinpanta0/Patient-Readmission-Prediction",
  "model_type": "sequence-classification",
  "library": {
    "random_forest": "scikit-learn",
    "logistic_regression": "scikit-learn",
    "k_nearest": "scikit-learn",
    "svc": "scikit-learn",
    "naive_bayes": "scikit-learn",
    "neural_network": "keras",
    "cross_validation_random_forest": "scikit-learn",
    "cross_validation_logistic_regression": "scikit-learn",
    "cross_validation_lightgbm": "LightGBM"
  },
  "model_architectures": {
    "random_forest": "RandomForestClassifier",
    "logistic_regression": "LogisticRegression",
    "k_nearest": "KNeighborsClassifier",
    "svc": "SVC",
    "naive_bayes": "MultinomialNB",
    "neural_network": "NeuralNetwork",
    "cross_validation_random_forest": "RandomForestClassifier",
    "cross_validation_logistic_regression": "LogisticRegression",
    "cross_validation_lightgbm": "LGBMClassifier"
  },
  "model_paths": {
    "random_forest": "model_RandomForestClassifier.pkl",
    "logistic_regression": "model_Logistic_Regression.pkl",
    "k_nearest": "model_K_nearest.pkl",
    "svc": "model_svc.pkl",
    "naive_bayes": "model_naive_bayes.pkl",
    "neural_network": "neural_network.keras",
    "cross_validation_random_forest": "model_rf.pkl",
    "cross_validation_logistic_regression": "model_lr.pkl",
    "cross_validation_lightgbm": "model_lgbm.pkl"
  },
  "model_classes": {
    "random_forest": "RandomForestClassifier",
    "logistic_regression": "LogisticRegression",
    "k_nearest": "KNeighborsClassifier",
    "svc": "SVC",
    "naive_bayes": "MultinomialNB",
    "neural_network": "NeuralNetwork",
    "cross_validation_random_forest": "RandomForestClassifier",
    "cross_validation_logistic_regression": "LogisticRegression"
  },
  "model_configs": {
    "random_forest": {
      "n_estimators": 100,
      "max_depth": 5
    },
    "logistic_regression": {
      "C": 1,
      "max_iter": 1000
    },
    "k_nearest": {
      "n_neighbors": 5
    },
    "svc": {
      "C": 1,
      "kernel": "linear"
    },
    "naive_bayes": {
      "alpha": 1
    },
    "neural_network": {
      "input_dim": 10,
      "output_dim": 1,
      "hidden_dim": 10
    },
    "cross_validation_random_forest": {
      "n_estimators": 100,
      "max_depth": 5
    },
    "cross_validation_logistic_regression": {
      "C": 1,
      "max_iter": 1000
    },
    "cross_validation_lightgbm": {
      "random_state": 42
    }
  }
}

metrics

Model Accuracy Precision Recall AUC-ROC
Random Forest 0.86544 0.8734358240972471 0.8337883959044369 0.8635809449401703
Logistic Regression 0.74736 0.7493540051679587 0.6928327645051194 0.7441573461079813
K-Nearest Neighbors 0.84112 0.8543724844493231 0.7969283276450512 0.838524404786381
Support Vector Classifier 0.84256 0.8492462311557789 0.8075085324232082 0.8405012541634113
Naive Bayes 0.74176 0.7692307692307693 0.6416382252559727 0.7358793535918418
Neural Network 0.87664 0.889009009009009 0.8419795221843004 0.8746042189234755
Random Forest (Cross-Validation) 0.86544 0.8734358240972471 0.8337883959044369 0.8635809449401703
Logistic Regression (Cross-Validation) 0.74736 0.7493540051679587 0.6928327645051194 0.7441573461079813
LightGBM (Cross-Validation) 0.8728 0.8773418168964299 0.847098976109215 0.8712904519100293
Random Forest Logistic Regression K-Nearest Neighbors Support Vector Classifier Naive Bayes Neural Network Random Forest (Cross-Validation) Logistic Regression (Cross-Validation) LightGBM (Cross-Validation)
1.0 0.7453866666666666 0.8901866666666667 0.8530133333333333 0.7455466666666667 0.88288 1.0 0.7453866666666666 0.9045866666666667
1.0 0.7449201741654572 0.9005328596802842 0.8556024378809189 0.7743332882090158 0.8964114832535885 1.0 0.7449201741654572 0.910874897792314
1.0 0.6979827742520399 0.8618540344514959 0.8272892112420671 0.6482320942883046 0.849274705349048 1.0 0.6979827742520399 0.8837262012692656
1.0 0.7427552396345833 0.8886139001594574 0.8515853672571407 0.7401446588709305 0.8810145438895148 1.0 0.7427552396345833 0.9034286859660855
Downloads last month
26
Inference Examples
Inference API (serverless) does not yet support keras models for this pipeline type.

Dataset used to train prabinpanta0/Patient-Readmission-Prediction