Hi ...
I'm having a dataset from 2011 to 2013. There are 500 data points. The dataset consists of two columns. One is the data column and other is the 'ExchangeRate' column. I was asked to divide the dataset into training and testing and requested to predict the data. may I kindly know whether there any possibility to use SVR for univariate data. What is the input that I might need to insert for the model, since it have a single column along with time. Kindly help
I suggest you follow a similar query discussed here https://www.researchgate.net/post/Is_there_a_SVM_R_code_to_forecast_univariate_time_series
Also Article A machine learning approach to univariate time series foreca...
check out this article
https://stackoverflow.com/questions/64198200/how-do-you-forecast-future-values-using-support-vector-regression-in-r
You can never directly do this with SVM actually. In case of Univariate data, you must use a grey box model instead of a black box model/ machine learning model. But if you want to use any machine learning model for predicting time series univariate data then you have to do some extra work. Generally differentiating the outputs successively will give you a better understanding and more input columns for predicting the output data. You can transfer the univariate data into different forms using different data transformation techniques and you can use them as input variables and then you will be able to use support vector regression for predicting this type of data.
Using a dynamic model like dynamic linear models or nonlinear models of differential equations are difference equations For forecasting the time series univariate data.
Otherwise you can use some machine learning methods like long short term memory which is one kind of recurrent neural network, this algorithm can predict and forecast the time-dependent data.
- Badges
- Science topic