Yapay Zeka Hızlı Bakış & Çıkarımlar
AI InsightAlthough performances indicators are very close to each other according to the internal indexes, it can be stated that the model obtained using the Manhattan distance measure and the k-Means algorithm has higher performance in terms of clustering success.
📌 3 Maddede Bu Makale
Abstract
With location-based smart applications, the flow of life can be facilitated and support can be provided in case of security and emergency situations. Indoor location detection provides various conveniences in complex structures such as hospitals, schools, shopping centers, etc. Indoor location detection studies are carried out by using data related to location and signal and machine learning methods. Machine learning has become frequently used as a solution method in this field, as in many other fields. When the studies in the literature are examined, it is seen that the studies are mainly focused on producing solutions with supervised machine learning algorithms. Unsupervised algorithms are frequently used to determine the labels of data groups that do not have labels. In this direction, it can be seen as the first step in labeling the data collected in indoor positioning studies and then using it for training predictive models to be developed with supervised learning methods. For this reason, the results to be obtained regarding the success and usefulness of cluster analysis will constitute an important basis for further studies. In this study, it is aimed to examine the success of unsupervised learning, in other words, clustering algorithms. The Wireless Indoor Localization Data Set and well-known k-Means and Fuzzy c-Means algorithms have been used with different distance measure. The obtained methods performances have been evaluated with internal and external indices. The results show that the clustering algorithms can cluster correctly data points in the range of 93-95% according to the accuracy and F measure value. Although performances indicators are very close to each other according to the internal indexes, it can be stated that the model obtained using the Manhattan distance measure and the k-Means algorithm has higher performance in terms of clustering success.