<![CDATA[At this time with the growing amount of information, the concept of data mining getting known as an important tool in the management information. Refers to the concept of data mining, the most popular concept in data mining is a clustering technique. One well known clustering method is k-means traditional. But in its application, k-means method has some problems such as determining the value of K cluster and determining the initial cluster centers were done randomly making process was inconsistent and the results of the cluster becomes worse. Therefore, there is a method to overcome these problems are improved semi-supervised k-means clustering. With improved semi-supervised method that combines the supervised and unsupervised method, users only need to label a bit of data that has not been labeled, then the labeled data is used to find the optimal value of initial cluster center and K cluster that will optimizes the process and result of clustering process. On implementation, this research combine k-means algorithm and improved semi-supervised k-means to clustering human development index (HDI) data. HDI data chosen because it has the right characteristics for clustering such amounts of data and the data is divided into several clusters. On the testing improved semi-supervised k-means method giving out the average accuracy of 90.3%, better than k-means clustering that giving 73.7% accuracy. In the second testing, improved semi-supervised k-means method produces an average time for one convergent 1222.9959 seconds, better than k-means with 1504.75 seconds. The third testing, improved semi-supervised k-means generates an average number of iterations for one convergent more efficient than k-means with the number of iterations of 7.11 compared 9.72. Last, on the cluster quality testing using silhouette coefficient, improved semi-supervised k-means method giving average value 0.69880, better than the traditional k-means with an average value of 0.62734.]]>
