TimeSeries forecasting not work well

I use the following data format for time series forecasting.The data is on the half-hour dimension,and the forecasted column is `NextTimeCount`.

Use data from the past 24 hours or more to predict the next hour data.
No matter how adjust the `windowSize` paramter, there is no good result.
```
Year,Time,RealCount,DropCount,TurnOnRate,ServiceRate,ProductRate,UsageRate,ShareRate,AHT,RealManCount,NextTimeCount,PreTimeCount,Month,Day,TotalCount,DayOfWeek
2020,0,130,4,0.97,0.96,0.72,0.96,0.75,3.97,26,116,0,9,1,134,2
2020,30,115,1,0.99,0.98,0.76,0.97,0.78,4.7,26,96,134,9,1,116,2
2020,60,84,12,0.86,0.85,0.7,0.95,0.74,3.83,15,61,116,9,1,96,2
2020,90,61,0,1,1,0.67,0.94,0.71,4.34,15,57,96,9,1,61,2
2020,120,53,4,0.92,0.92,0.81,0.96,0.85,4.63,10,54,61,9,1,57,2
```
Even the predictions in the [example](https://github.com/dotnet/machinelearning-samples/tree/master/samples/csharp/end-to-end-apps/Forecasting-Sales) are not very accurate,The prediction number is only half of regression model.

Any suggestion?