update to the NRQL predictions and PREDICT syntax docs

src/content/docs/nrql/nrql-syntax-clauses-functions.mdx

@@ -785,32 +785,128 @@ As noted in our [basic NRQL syntax doc](/docs/query-your-data/nrql-new-relic-que
   We're still working on this feature, but we'd love for you to try it out!
 
   This feature is currently provided as part of a preview program pursuant to our [pre-release policies](/docs/licenses/license-information/referenced-policies/new-relic-pre-release-policy).
-</Callout>
+  </Callout>
 
     ```sql
     FROM Transaction SELECT count(*) WHERE error IS TRUE TIMESERIES PREDICT
     ...
     ```
 
-    With the `PREDICT` clause, you can incorporate predictions regarding future data trends into line charts based on historical data. This clause:
+    With the `PREDICT` clause, you can add predictions of future data trends into line charts based on historical data. Consider the following points when using this clause:
 
     * Applies only to queries with a `TIMESERIES` clause.
     * Uses the `TIMESERIES <time period>` as the interval for predicted data points.
-    * Sets the prediction range to 20% of the query window by default.
-    * Processes historical data from the current query window and the two preceding ones to generate predictions by default.
     * Does not support [metric timeslice data](/docs/data-analysis/metrics/analyze-your-metrics/data-collection-metric-timeslice-event-data#timeslice-data) while in Public Preview.
 
     <Callout variant="important">
       The `PREDICT` clause is not supported when used with the [`COMPARE WITH`](/docs/nrql/nrql-syntax-clauses-functions/#sel-compare) clause or [nested aggregations](/docs/nrql/using-nrql/nested-aggregation-make-ordered-computations-single-query) inside [subqueries](/docs/nrql/using-nrql/subqueries-in-nrql/).
     </Callout>
-    You can additionally append `PREDICT <model>`, `BY <time range>`, and `USING <time range>` to refine the prediction trend to your specific needs. 
-    To indicate the time range and historical data span, use `integer units`.
 
-            ```sql
-        FROM Transaction SELECT count(*) WHERE error IS TRUE TIMESERIES PREDICT holtwinters(beta: 0.5) BY 3 hours USING 1 day
+    ### Customization [#customization]
+    To learn about the default behaviour of the prediction model, refer to [NRQL Predictions](/docs/query-your-data/explore-query-data/use-charts/nrql-predictions/#defaults). You can customize the predictions by manually setting the training model's hyperparameters and adding the `USING` and `BY` keywords.
+
+    <CollapserGroup>
+      <Collapser id="hyperparameters" title="Set model hyperparameters.">
+      Currently, <DNT>**Holt-Winters**</DNT> is the only prediction algorithm we support, and it comes with several hyperparameters to help you fine-tune your predictions. Refer to the following table to understand the behavior of each hyperparameter and choose the ones that suit your use case.
+
+      To modify your query, append `holtwinters(<hyperparameters>: <value>)` after the `PREDICT` clause.
+
+       <table id="hyperparameters">
+        <thead>
+          <tr>
+            <th colSpan="3">
+            </th>
+      </tr>
+      <tr>
+        <th>Hyperparameter</th>
+        <th>Value</th>
+        <th>Behavior</th>
+      </tr>
+    </thead>
+
+        <tbody>
+          <tr>
+            <td rowspan="3" style="width: 300px;">`seasonality`</td>
+            <td>`<season length>`</td>
+            <td>
+              Sets the seasonality of the time series to a specific season length. Make sure the season length is an `integer unit` of time.  
+              **Examples**: `1 day` or `3 hours`
+            </td>
+          </tr>
+          <tr>
+            <td>`AUTO`</td>
+            <td>
+              Automatically detects the seasonality of the time series and generates predictions with the identified seasonality.
+            </td>
+          </tr>
+          <tr>
+            <td>`NONE`</td>
+            <td>
+              Disables the seasonality of the time series and generates a non-seasonal prediction.
+            </td>
+          </tr>
+
+          <tr>
+            <td>`alpha`</td>
+            <td>0 to 1</td>
+            <td>
+            Sets the level smoothing factor.
+            * **Higher value**: More weight on recent data.
+            * **Lower value**: More weight on historical data.
+            </td>
+          </tr>
+
+          <tr>
+            <td>`beta`</td>
+            <td>0 to 1</td>
+            <td>
+            Sets the trend smoothing factor.
+            * **Higher value**: More weight on recent data.
+            * **Lower value**: More weight on historical data.
+            </td>
+          </tr>
+
+          <tr>
+            <td>`gamma`</td>
+            <td>0 to 1</td>
+            <td>
+            Sets the season smoothing factor.
+            * **Higher value**: More weight on recent data.
+            * **Lower value**: More weight on historical data.
+            </td>
+          </tr>
+
+          <tr>
+            <td>`phi`</td>
+            <td>0.98 to 1</td>
+            <td>
+            Sets the trend damping parameter.
+            * **Lower value**: More damping effect to the prediction curve that gradually levels out to a flat curve.
+            * **Higher value**: Lesser damping effect to the prediction curve.
+            </td>
+          </tr>
+        </tbody>
+      </table>  
+      </Collapser>
+
+      <Collapser id="by" title={<>Add the <InlineCode>BY</InlineCode> keyword.</>}>
+      Use the `BY` keyword to set how far ahead in time you want the model to predict. For example, `PREDICT BY 3 hours` generates a prediction 3 hours into the future from the latest point in the time series. Make sure, the time span is an `integer unit` of time.
+      </Collapser>
+
+      <Collapser id="using" title={<>Add the <InlineCode>USING</InlineCode> keyword.</>}>
+      Use the `USING` keyword to specify the amount of historical data you want to use to train the model. For example, `PREDICT USING 1 day` uses the query window along with the data from the preceding day to train the model. Make sure the time span is an `integer unit` of time.
+      amount of time.
+      </Collapser>
+    </CollapserGroup>
+
+
+    Example query using hyperparameters, the `BY` keyword, and the `USING` keyword.
+
+        ```sql
+        FROM Transaction SELECT count(*) WHERE error IS TRUE TIMESERIES PREDICT holtwinters(seasonality: AUTO, alpha: 0.2) BY 1 hour USING 2 hours
         ```
 
-  To learn more about how and when you can use `PREDICT`, refer to [NRQL Predictions](/docs/query-your-data/explore-query-data/use-charts/nrql-predictions/).
+    To learn more about how and when you can use `PREDICT`, refer to [NRQL Predictions](/docs/query-your-data/explore-query-data/use-charts/nrql-predictions/).
   </Collapser>
 
   <Collapser