It is not exactly an answer to what I said. Yes it is 1D(univariate time series), but it has some other properties; this is why we use sequential models when dealing with time series data.

Splitting time series data as you did is totally wrong and misleading. You should use forward chaining. Please take a look at this links.

Using k-fold cross-validation for time-series model selection

Time Series Nested Cross-Validation