I am researching methods to increase the accuracy of an algorithm that is currently used to analyse radiation patterns as they hit our sensor. (For the non-physicists, we will mainly see Alphas, Betas and Gammas which are the 3 most common radiation particles)
The 3 types of radiation make very distinctly different patterns on our sensor (which can be viewed graphically) which we call clusters (clusters of activated pixels). I can retrieve lots of data about every cluster such as: Density, Roundness, Linearity, (and any others that you may think of?)
But my question is: With a set of clusters from previous data sets that I can assign each as a Alpha, Beta or Gamma, is it possible to write an algorithm/equation which gives the most likely type particle for new clusters?
For example here is some sample data (that ive made up on the spot, for the real algorithm I would try to use more like 500 pre-classified clusters each with around more like 6 variables each rather than the 3 shown here):
Roundness = 0.9, Density = 0.8, Radius = 4, Cluster Type = Alpha
Roundness = 0.5, Density = 0.2, Radius = 8, Cluster Type = Beta
Roundness = 0.3, Density = 0.4, Radius = 7, Cluster Type = Beta
Roundness = 1, Density = 1, Radius = 1, Cluster Type = Gamma
With this data, could we go about forming a method for determining the type of cluster that has the variables: Roundness = 0.7, Density = 0.3, Radius = 6?
If so, what or where should I research to learn more about such things?
I believe this is relevant in fitting algorithms and scalar fields but I am not certain as im still doing pre-university maths.
Any help in the right direction would be most appreciated, Thank You!