python LinearRegression 进行实时预测

Question

i received data from many captors I want to predict electronic failure with linear regression in real time (i want to add new values to my model)

i have small example :

from sklearn.linear_model import LinearRegression
from sklearn.datasets import make_regression
#dataset
#X : Date in millisecond; temperature degree; humidity %
#y : 0= no problem; 1 = electronic failure
X=[[969695100000,15,10],[969788280000,30,50],[975042120000,20,3]]
y=[0,1,0]
model = LinearRegression()
model.fit(X, y)


# Prediction
Xnew, _ = make_regression(n_samples=3, n_features=3, noise=0.1, random_state=1)
ynew = model.predict(Xnew)
for i in range(len(Xnew)):
    print("X=%s, Predicted=%s" % (Xnew[i], ynew[i]))

i have real data X : Date; temperature and humidity and y, 0 => no problem, 1 the sensor has failed I have new data every days, i want to update my model every day.

my goal is with this data to predict a sensor failure by tomorrow.

my question is: how add data to my model ?

Answer 1

i found solution for update model in realtime, i use partial_fit, i updated my code like this :

import numpy as np
from sklearn import linear_model
from sklearn.datasets import make_regression

n_samples, n_features = 10, 5

X=[[969695100000,15,10],[969788280000,30,50],[975042120000,20,3]]
y=[0,1,0]
model = linear_model.SGDRegressor()
for i in range(0,1000):
    model.partial_fit(X, y)

Xnew, _ = make_regression(n_samples=3, n_features=3, noise=0.1, random_state=1)
ynew = model.predict(X)
for i in range(len(Xnew)):
    print("X=%s, Predicted=%s" % (Xnew[i], ynew[i]))

i have last question, it's possible to predict y value with only date ? actualy for predict y tomorow i need to have all X data, it's possible to predict with only : X=[975042120100] without temperature and humidity, only date in millisecond ?