Using PSO and RST to Predict the Resistant Capacity of Connections in Composite Structures

An analysis about the measure quality of similarity and its applications in machine learning

“…The employment of this type of similarity function is studied in [13] to give solution to the problem of the prediction of the resistant capacity of stud connectors in composite structures using the k-NN method like functions approximation. The following alternatives were studied to calculate the weight: assigning the same weight to each feature, three alternatives based in the expert criteria and the weights calculated by the maxQS method.…”

“…Subsequently, it is analyzed the solution proposed in [13,14] to build similarity relations. The similarity relation is formulated as xRy ⇔ F (x, y) ≥ ε, x, y ∈ U , and F (x, y) is a similarity function with values in [0,1].…”

“…To overcome this restriction the similarity quality measure is introduced in [17], this measure is defined according to a similarity relation, which allows working with mixed data to describe the objects and working with discrete or continuous decision features. This measure can be analyzed as a combination of the granularity of the universe according to CG and DG, and the principle of the Case-based reasoning.…”

“…The new problem is finding the appropriate similarity relation for each application domain. A method to build similarity relations based on the similarity quality measure is proposed in [13,14], this includes to compute the weights for the features.…”

“…This is an optimization problem that can be solved using a heuristic search, in which the heuristic evaluation function defined by the similarity quality measure is maximized. In [13,14], the Particle Swarm Optimization (PSO) [28] has been used to solve the problem. In this paper (to see section 4.1) an extension of the study is presented by using as heuristic method the Univariant Marginals Distribution Algorimth (UMDA), instead of PSO.…”