# Doubt regarding polynomial regression in temprature sensing

I am having a doubt about polynomial coefficient and frame size.if we set polynomial coefficient be 5 it means degree of 5.So a polynomial of degree 5 would be created after some data point that we will be collecting.So if we get polynomial of that data, why we want frame size ?
please reply if am interpreting something wrong.

Frame size is the number of previous data point which use to predict the trend of data by visualizer.

For explaining this, I would like to first explain about how Machine Learning Algorithm works:
Machine Learning Algorithm creates a models ( in present scenario it is polynomial function) and trains it using the training data which in this case is the frame size. It creates a cost function (like difference in values what we are getting from the polynomial function and what actually is) and it tries to minimize it as much as possible by altering the model( which in present case is the degree of the polynomial and itâ€™s coefficient). So, the frame size and degree of polynomial is completely different things. Frame size is the training data whereas the degree of polynomial is for creating a model.
Now, whatâ€™s the need of altering the frame size?
Answer is that the most common problem which arises during running a model is: 1. Overfitting 2. Underfitting
Underfitting occurs when a statistical model or machine learning algorithm cannot capture the underlying trend of the data. In other words, when the model or the algorithm does not fit the data well enough.
Overfitting occurs when a statistical model or algorithm captures the noise of the data. In other words, when the model or the algorithm fits the data too well.

So, in present scenario, if we increase the degree of polynomial more than required then overfitting occurs. To minimize overfitting we increase the frame size or reduce degree of ploynomial
And if we decrease the degree of polynomial more than required then underfitting occurs. To minimize it we increase the degree or reduce frame size.
Hope it will help you in understanding the concept.

3 Likes

in the bolt cloud visualizer, polynomial coefficient value is remain fixed.So how the model is altering the value ? i simply don.t able to understand this thing ? @priyadarshi.pritam12

Hi @namdev.1:

Please clarify, why do you feel that the polynomial coefficient value remains fixed?

For the visualiser, you are setting the number of polynomial coefficients that will be used to generate the polynomial function, and frame size is the number of data points that the visualiser uses to generate the function.

For example, if you set the frame size to 10, and polynomial coefficient to 3, for the following data series:

99,46,48,19,99,47,11,24,92,26,94,71,45

First the visualisers will use the first 10 data points which are â€ś99,46,48,19,99,47,11,24,92,26â€ť, to generate a polynomial function, with degree 3, and it will use the function to generate a prediction history data point, for the time at which the data point 94 (11th data point) was collected.
Then the visualiser will repeat the process for the data points â€ś46,48,19,99,47,11,24,92,26,94â€ť (data point 2 to 11) to generate a new polynomial function of degree 3and get a prediction history data point for the time at which the data point 45(12th data point) was collected.
It will similarly repeat the process for data points 3 to data point 12, and get the equivalent prediction history data point for the 13th data point.

Finally, the visualiser will use the data points 4 to 13 to generate a polynomial function with degree 3, and use the function to generate the future data points. The number of future data points that the system predicts is the number that you set as the Prediction points in the visualiser.

I hope this clears your doubt.

3 Likes

thank you so much @vinayak.joshi for clarifying my doubt.

1 Like

@namdev.1 I think now you might have understood my explainationâ€¦For any furthur queries related to this feel free to ask