# Feature extraction in time series data as input variables for machine learning algorithms

I have worked with time series data to predict the defect in production lines. I want to extract the feature from time series data as inputs variables for machine learning algorithms (such as support vector machine). My dataset looks like:

Timestamp  Pressure
t0           x0
t1           x1
t2           x2
.             .
.             .
.             .
tn           xn


There is a threshold b. If xt>b that means the defect happened. My goals are to extract features from time series dataset above and put it into algorithms to predict the value of time stamp tn+1, tn+2 (short-term) and long-term tn+10. However, I have not yet found the way to extract feature and how to bring the threshold into the algorithm. Could anyone suggest me how to deal with it? Thank you.