I am new to machine learning techniques. I was going through few supervised machine learning model examples and i have doubt in predicting future values.

I have daily time series dataset from database where my target variable is complete noise signal like this:

To train and predict the models, we divide the dataset into train set, validation set and test set to check if the model is efficient or not. I have two independent variable and one target variable. I am using linear regression,Keras LSTM and other models.

My basic question is how do i predict for future values (for next one week, one month) when i don't have independent variables for next week? what am i supposed to give as predictors in that case?

Any information is much appreciated.


1 Answer 1


Welcome to the site!

Firstly, when we use any kind of predicting algorithm, then you need to have the future values of the independent variables(features for explaining the target variable), for Prediction.

Secondly, If you don't have such values then you cannot use predicting algorithms, we have forecasting algorithms to do such task. To apply this you need to have data which is time series data. From the sample and graph I think your data is time series data.

FYI, you can also use additional features even in time series. For Feature Engineering on Time Series data you can go through the link.

If you need any additional information let me know.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.