The application of a structural equation modeling (SEM) assumes that all data follow only one model. This assumption may be inaccurate in certain cases because individuals tend to differ in their responses, and failure to consider heterogeneity may threaten the validity of the SEM results. This study focuses on unobservable heterogeneity, where the difference between two or more data sets does not depend on observable characteristics. In this study, we propose a new method for estimating SEM parameters containing unobserved heterogeneity within the data and assume that the heterogeneity arises from the outer model and inner model. The method combines partial least squares (PLS) and modified fuzzy clustering. Initially, each observation was randomly assigned weights in each selected segment. These weights continued to be iteratively updated using a specific objective function. The sum of the weighted residual squares resulting from the outer and inner models of PLS-SEM is an objective function that must be minimized. We then conducted a simulation study to evaluate the performance of the method by considering various factors, including the number of segments, model specifications, residual variance of endogenous latent variables, residual variance of indicators, population size, and distribution of latent variables. From the simulation study and its application to the actual data, we conclude that the proposed method can classify observations into correct segments and precisely predict SEM parameters in each segment.

Credit Scoring is designed so that lenders can easily make decisions regarding whether a loan proposal from a prospective customer is worthy of approval or not. This study examines the application of the Multi Local Means Based K Harmonic Nearest Neighbor (MLMKHNN) method in the case of motorcycle credit in a financial institution. The classification capability of this method in detecting potential borrowers into the credit category is either good or bad compared to its previous method, Local Means Based K Harmonic Nearest Neighbor (LMKNN). In this case the MLMKHNN method has not shown better performance than the LMKNN method. At the same level of total accuracy, MLMKHNN requires more numbers of neighbors than the number of neighbors required by the LMKNN method. Keywords: sampling design, all possible samples, statistical efficiency, cost efficiency

Credit scoring is one of the crucial task and a core responsibility for financial institutions in their risk management. This study aims to apply the pseudo nearest neighbour (PNN) method as a tool to identify which prospective borrowers are eligible for their loan proposals. If a new borrower has characteristics closer to a good historical borrower then the loan proposal is worthy to approval. But if not, the proposed loan will be refused. The historical data in this paper are credit data from a national bank in Indonesia. The characteristics of historical debtors consist of age, amount of a child, length time of business, income, loans amount, and the period of credit. The best classification of k-NN is using k = 1, because it makes the smallest error 1,89%. While the best classification of PNN is using k = 13 with the smallest error 20,75%. Based on total accuracy of classification shows that the credit classification of debtors using k-NN is more appropriate than PNN.

In structural equation modeling (SEM), it is usually assumed that all observations follow only one model. This becomes irrelevant if the observations contain natural groups, each of which has a different SEM model. Mukid et al (2002) have proposed the partial least squares-modified fuzzy clustering method (PLSMFC) as a way to find groups of observations and at the same time estimate the parameters of the SEM model. This research aims to understand the performance of the PLSMFC method in finding groups of observations characterized by different forms of structural equation models. The goal was achieved by conducting a simulation study involving factors such as SEM model specification and number of clusters. The procedure used is to force the generated data into a different number of segments. The segment validity measures used are the fuzziness performance index (FPI) and normalized classification entropy (NCE). The correct number of segments is indicated by the smallest FPI and NCE values. Based on simulation studies, it is known that the PLSMFC method can detect segments accurately, especially if the size of the segments used to reallocate observations is larger than the number of segments used to generate the data.

The queuing system is a set of customers, services and a set of rules governing the arrival of its customers and services. Queue is a waiting phenomenon that is part of everyday human life. The queue is formed if the number of subscribers to be served exceeds the available service capacity. Queue phenomenon one of them seen in the queue nonpatas buses at Terminal Tirtonadi Surakarta. Nonpatas bus lanes studied include non-purpose buses Surabaya, Karanganyar, Wonogiri, Purwodadi and Pedesaan. The queue displant used is FIFO (First In First Out). For the five nonpatas bus lanes it meets steady state conditions because it has utility value less than 1. The selected model is a model that has the following 4 types of distributions: Erlang, Weibull, Gamma and Lognormal. The queue model generated for the five tracks (ERLA/ERLA/1):(GD/∞/∞) for Surabaya nonpatas buses, (ERLA/WEIB/1):(GD/∞/∞) for Karanganyar nonpatas buses, (GAMM/WEIB/1):(GD/∞/∞) for Wonogiri nonpatas buses, (ERLA/WEIB/1):(GD/∞/∞) for Purwodadi nonpatas buses, (WEIB/LOGN/1):(GD/∞/∞) for Pedesaan nonpatas buses. Based on the value of the system performance measure indicated that the five lines are queue system is good. Keywords: Beta, Erlang, FIFO, Gamma, Steady State Conditions, Lognormal, Queue Model, Queuing Systems, System Performance Measure, Weibull