Michel chamat dimitrios bersi kodra

Collected
Data
Training
Set
Test
Set
Training
Model
Training
Database
Prediction
Output
Machine Learning

---

42 
Fig. 20. Neural network algorithm for MCS prediction. 
In details, the future MCS prediction, output, is as follows: 
● Predict the scenario (hidden node) based on the input data 
● Predict the MCS based on output of the hidden node 
By choosing the closest UE scenario from the training dataset, using the 
parameters mentioned in table 3 as inputs, the future MCS can be predicted 
as a smaller yet more accurate and relative dataset will be used.  
The flowchart in Fig. 21 shows the detailed steps taken to make an MCS 
prediction based on the input data from the UE. 
SINR
MCS
Data
Size
HARQ
Scenario
1
Scenario
10
Scenario
20
MCS
1
MCS
14
MCS
28

---

43 
Fig. 21. Flow chart of the MCS prediction. 
The UE training dataset is a sub-dataset of the main training dataset, and it 
is selected using the predicted UE scenario (hidden node in the neural 
network). The second prediction follows using the same input data and the 
selected sub-dataset, so no additional data is added to the overall prediction.
 
 
 
Input
Data
MCS
Prediction
Prediction
Prediction
Training
Dataset
UE Training