'How to perform a split on data when you depend on dates? (python)

I'm working on a project where I predict the value of a cryptocurrency the next day. The data provided has the Date and Price of each respective currency. The problem is, since the data refers back to 2017, some cryptocurrencies didn't exist there yet. So, I can't make predictions of a cryptocurrency, considering a data where it didn't exist. How should I split (train/test) my data? Or should I separate each cryptocurrency into a different dataset? Would that be efficient taking into account I have to automate my model as much as possible? Thank you.

time-seriesdata-scienceforecastingcryptocurrencytrain-test-split


Sources

This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.

Source: Stack Overflow

Solution Source

Related Questions

Can I specify the dates and times of a time series in R?

Trend and Seasonality from time series

Error installing FBPROPHET in my windows machine

AttributeError: module 'tensorflow.python.pywrap_tensorflow' has no attribute 'TFE_Py_RegisterExceptionClass'

Error in modelinc[15] = dim(variance.model$external.regressors)[2] : replacement has length zero

Unable to deploy fbprophet in heroku

Two methods of recovering fitted values from a Bayesian Structural Time Series model yield different results

Filter prometheus gauge metrics by value range

Slice DataFrame using Dates from a List and its offset as a range for slicing in a for loop

In Seasonal Decompose Seasonal plot seems off

Pandas TimeSeries resample produces NaNs

Plotly: How to highlight weekends without looping through the dataset?

multivariate xgboost time series

Compute annual mean using x-arrays

Is there a way to forecast sales for multiple products across multiple stores?