docs: added time series validation to features section of website (#565)

LGTM. Thanks @michellegrushkometa and congrats for your first official PR merged to main branch! <3

website/docs/features2.mdx

@@ -230,7 +230,6 @@ Ridge regression has an additional benefit of being relatively easy to interpret
 
 ##  Multi-Objective Hyperparameter Optimization with Nevergrad
 
----
 
 One of the most important innovation in Robyn is the implementation of multi-objective hyperparameter optimization in MMM. This enables us to automate the selection of adstocking, saturation, regularization penalty and even training size for time-series validation. At the same time, the ability to optimise towards multiple "goals", implemented as objective functions, provides us the edge to produce model candidates with great predictive power as well as more realistic effect decomposition.
 
@@ -281,6 +280,16 @@ To investigate the convergence of your multi-objective optimization. See below f
 <img alt="moo distrb plot" src={useBaseUrl('img/moo_distrb_plot.png')} />
 <img alt="moo cloud plot" src={useBaseUrl('img/moo_cloud_plot.png')} />
 
+---
+## Time Series Validation
+
+When `ts_validation = TRUE` in `robyn_run()` a 3-way-split time series for NRMSE validation is enabled.
+A time series validation parameter `train_size` is included as one of Robyn's hyperparameters. When `ts_validation = TRUE` in `robyn_run()`, `train_size` defines the percentage of data used for training, validation and out-of-sample testing. For example, when `train_size = 0.7`, `val_size` and `test_size` will be 0.15 each. This hyperparameter is customizable or can be fixed with a default range of `c(0.5, 0.8)` and must be between `c(0.1, 1)`.
+
+<img alt="time series validation" src={useBaseUrl('img/time_series_validation_and_convergence.png')} />
+
+<img alt="time series validation" src={useBaseUrl('img/actual_vs_predicted_response_ts.png')} />
+
 ---
 
 ## Calibration with Experiments