SAS Visual Data Mining and Machine Learning in SAS Viya offers a number of algorithms for training powerful predictive models, such as gradient boosting, forest, and deep learning models. Although these models are powerful, they are often too complex for people to understand by directly inspecting the model parameters. The “black-box” nature of these models limits their use in highly regulated industries such as banking, insurance, and health care. This paper introduces various model-agnostic interpretability techniques available in SAS Viya that enable you to explain and understand machine learning models. Methods include partial dependency (PD) plots, independent conditional expectation (ICE) plots, local interpretable model-agnostic explanations (LIME), and Shapley values. This paper introduces these methods and demonstrates their use in two scenarios: a business-centered modeling task and a health-care modeling task. Also shown are the two different interfaces to these methods in SAS Viya: Model Studio and the SAS Viya programming interface.

The support vector machines (SVM) algorithm is a popular classification technique in data mining and machine learning. In this paper, we propose a distributed SVM algorithm and demonstrate its use in a number of applications. The algorithm is named high-performance support vector machines (HPSVM). The major contribution of HPSVM is two-fold. First, HPSVM provides a new way to distribute computations to the machines in the cloud without shuffling the data. Second, HPSVM minimizes the inter-machine communications in order to maximize the performance. We apply HPSVM to some real-world classification problems and compare it with the state-of-the-art SVM technique implemented in R on several public data sets. HPSVM achieves similar or better results.

A wide variety of clustering algorithms are available, and there are numerous possibilities for evaluating clustering solutions against a gold standard. The choice of a suitable clustering algorithm and of a suitable measure for the evaluation depends on the data type; whether separate class label information exists (supervised clustering); and on the particular distribution of the observations, including characteristics such as the number of clusters, separability of the clusters, and the shape, size, and density of the clusters. This paper provides a survey of some of the most widely used clustering evaluation criteria. In addition, the paper describes recently developed criteria that are applicable for mixed interval-categorical data and for non-Euclidean distance metrics. Notable examples of the methods covered include residual sum-of-squares, purity, the silhouette measure, the Calinski-Harabasz measure, class-based precision and recall, the normalized mutual information, variation of information, and graph-sensitive indices.

Colorectal cancer (CRC) is the third most common cancer and the second leading cause of cancer-related deaths worldwide. Most CRC deaths are the result of progression of metastases. The assessment of metastases is done using the RECIST criterion, which is time consuming and subjective, as clinicians need to manually measure anatomical tumor sizes. AI has many successes in image object detection, but often suffers because the models used are not interpretable, leading to issues in trust and implementation in the clinical setting. We propose a framework for an AI-augmented system in which an interactive AI system assists clinicians in the metastasis assessment. We include model interpretability to give explanations of the reasoning of the underlying models.

Decision tree algorithms are very popular in the field of data mining. This paper proposes a distributed decision tree algorithm and shows examples of its implementation on big data platforms. The major contribution of this paper is the novel KS-Tree algorithm which builds a decision tree in a distributed environment. KS-Tree is applied to some real world data mining problems and compared with state-of-the-art decision tree techniques that are implemented in R and Apache Spark. The results show that KS-Tree can achieve better results, especially with large data sets. Furthermore, we demonstrate that KS-Tree can be applied to various data mining tasks, such as variable selection.

We explore the geometrical interpretation of the PCA based clustering algorithm Principal Direction Divisive Partitioning (PDDP). We give several examples where this algorithm breaks down, and suggest a new method, gap partitioning, which takes into account natural gaps in the data between clusters. Geometric features of the PCA space are derived and illustrated and experimental results are given which show our method is comparable on the datasets used in the original paper on PDDP.

Data clustering has various applications in a wide variety of fields ranging from social and biological sciences, to business, statistics, information retrieval, machine learning and data mining. Clustering refers to the process of grouping data based only on information found in the data which describes its characteristics and relationships. Although humans are generally very good at discovering patterns and classifying objects, clustering algorithms are able to discern similarities in data even when humans are not [6]. The main focus of our research has been document clustering, but we will demonstrate that our methods also work nicely on scientific data. In this paper, we propose an adaptation of the clustering algorithm known as Principal Direction Divisive Partitioning (PDDP) developed by Daniel Boley in [2] which is based Principal Components Analysis (PCA). PCA involves the eigenvector decomposition of a data covariance matrix, or equivalently a singular value decomposition (SVD) of a data matrix after mean centering. The name of our adaptation, Principal Direction Gap Partitioning (PDGP), borrows most of its name from PDDP as it follows many of the same steps that PDDP follows. The word “gap” replaces the word “divisive” in reference to how the algorithm splits data along natural gaps at each step. This concept will be further developed in the following sections, but it should be noted that PDGP is still a divisive algorithm in the same way that PDDP is.

Cluster Analytics helps to analyze the massive amounts of data which have accrued in this technological age. It employs the idea of clustering, or grouping, objects with similar traits within the data. The benefit of clustering is that the methods do not require any prior knowledge of the data. Hence, through cluster analysis, interpreting large data sets becomes, in most cases, much easier. However one of the major challenges in cluster analytics is determining the exact number of clusters, k, within the data. For methods such as k-means and nonnegative matrix factorization, choosing the appropriate k is important. Other methods such as Reverse Simon-Ando are not as dependent on beginning with the correct k. In this paper, we discuss these methods and apply them to several well-known data sets. We then explore techniques of deriving the number of clusters from the data set and lastly several points of theoretical interest.

We perform importance sampling for a randomized matrix multiplication algorithm by Drineas, Kannan, and Mahoney and derive probabilities that minimize the expected value (with regard to the distributions of the matrix elements) of the variance. We compare these optimized probabilities with uniform probabilities and derive conditions under which the actual variance of the optimized probabilities is lower. Numerical experiments with query matching in information retrieval applications illustrate that the optimized probabilities produce more accurate matchings than the uniform probabilities and that they can also be computed efficiently.