Basic Logistic Regression Model for Predictive Maintenance Problem

logistic_regression_training.py

import pickle
from sklearn.linear_model import LogisticRegression
import pandas
import json

from sklearn import preprocessing
from sklearn import metrics
from sklearn.model_selection import train_test_split


with open('training_data.json') as json_data:
    data = json.load(json_data)

df = pandas.DataFrame(data)

X = df[['temp', 'vibration', 'current', 'noise']]


def normalize_features(X):
    transformer = preprocessing.Normalizer().fit(X)
    return transformer.transform(X).tolist()


def encode_labels(y):
    enc = preprocessing.LabelEncoder()
    enc.fit(y)
    return [enc.transform(y).tolist(), enc.classes_.tolist()]


X = normalize_features(X)
y, encoding = encode_labels(df[['state']])

train_X, val_X, train_y, val_y = train_test_split(X, y, random_state=1)

regressor = LogisticRegression(solver='liblinear')
regressor.fit(train_X, train_y)

y_pred = regressor.predict(val_X)
accuracy = metrics.accuracy_score(y_pred, val_y)

print(accuracy)

file_handler = open('factory_linear_regression.pkl', 'wb')
payload = {"model": regressor, "encoding": encoding}
pickle.dump(payload, file_handler)
file_handler.close()

print("pickled model")