Colm23/machine-failure-prediction

1 2import joblib3 4from sklearn.datasets import fetch_openml5 6from sklearn.preprocessing import StandardScaler, OneHotEncoder7from sklearn.compose import make_column_transformer8 9from sklearn.pipeline import make_pipeline10 11from sklearn.model_selection import train_test_split, RandomizedSearchCV12 13from sklearn.linear_model import LogisticRegression14from sklearn.metrics import accuracy_score, classification_report15 16dataset = fetch_openml(data_id=42890, as_frame=True, parser="auto")17 18data_df = dataset.data19 20target = 'Machine failure'21numeric_features = [22    'Air temperature [K]',23    'Process temperature [K]',24    'Rotational speed [rpm]',25    'Torque [Nm]',26    'Tool wear [min]'27]28categorical_features = ['Type']29 30print("Creating data subsets")31 32X = data_df[numeric_features + categorical_features]33y = data_df[target]34 35Xtrain, Xtest, ytrain, ytest = train_test_split(36    X, y,37    test_size=0.2,38    random_state=4239)40 41preprocessor = make_column_transformer(42    (StandardScaler(), numeric_features),43    (OneHotEncoder(handle_unknown='ignore'), categorical_features)44)45 46model_logistic_regression = LogisticRegression(n_jobs=-1)47 48print("Estimating Best Model Pipeline")49 50model_pipeline = make_pipeline(51    preprocessor,52    model_logistic_regression53)54 55param_distribution = {56    "logisticregression__C": [0.001, 0.01, 0.1, 0.5, 1, 5, 10]57}58 59rand_search_cv = RandomizedSearchCV(60    model_pipeline,61    param_distribution,62    n_iter=3,63    cv=3,64    random_state=4265)66 67rand_search_cv.fit(Xtrain, ytrain)68 69print("Logging Metrics")70print(f"Accuracy: {rand_search_cv.best_score_}")71 72print("Serializing Model")73 74saved_model_path = "model.joblib"75 76joblib.dump(rand_search_cv.best_estimator_, saved_model_path)77