Predictive Modeling of Compressive Strength composition values for structural studies using K-medoids clustering and Quantile Regression Forests

The determination of structural strength and reliability begins from its precise concrete composition mixture and appropriate material selection. Concrete is the primary entity in the structural industry, and requires competent skills and expertise to understand and attain an ideal compressive strength. However, in most instances nowadays, innovative components such as plasticizer, fly ash, and blast furnace slag have been found to contribute to its compactness and promote the cementitious property of the compound. This paper presents the combination of clustering and regression technique to deliver comparable values that will help in the conceptualization of insights from visualized results. By the aid of a public dataset comprising of 1030 samples, the data points of the dataset will be categorized and the resulting compressive strength will be classified and forecasted through the use of k-medoids clustering and quantile regression forest. Clustering through k-medoids minimizes noise and outliers as compared to the conventional k-means. Identification of medoids in every cluster promotes averaging of points in terms of similarity and dissimilarity. Overall, an optimal prediction interval is generated for all observations in predicting compressive strength. The 90% interval limits was obtained by predicting the 5^th and 95^th percentile.