I have run a K-Nearest Neighbor on some sleep data I am using for a project. I was able to run the model to about a 45% accuracy. What I am struggling to understand is that I graphed out the accuracy of my model based on the K number. After K > 36 the accuracy plateaus and never again varies. I will admit that the correlation between my data is not strong, but does anyone know why this plateau would occur and no longer decrease?
To give some background, I am using exercise data along with my sleep data and determining if the time of day that I exercise impacts my sleep quality. There is some correlation to late workouts and poorer sleep but otherwise values in the afternoon and morning can vary greatly and don't seem to be a strong impact.
The Y variable in analysis = Sleep Quality factor
The X variable in analysis = The hour of day of exercise.
Graph: x = K number y = Accuracy
