عنوان مقاله
کاربرد نگاشتهای تکاملی فازی برای پیش بینی بلند مدت سرطان پروستات
فهرست مطالب
مقدمه
نگاشت شناختی فازی – پیش زمینه نظری
پیش زمینه پزشکی
یادگیریFCM با قدرت پیش بینی بلند مدت
آزمایش های محاسباتی
نتیجه گیری
بخشی از مقاله
نگاشت شناختی فازی – پیش زمینه نظری
نگاشت های شناختی فازی یک توسعه ای از نگاشت های شناختی را تشکیل می دهند که دارای ویژگی های اصلی منطق فازی وشبکه های عصبی هستند.این نگاشت ها برای اولین بار توسطkosko به عنوان گراف های جهت دار مثبت برای رائه استدلال علی و پردازش استنباط محاسباتی ، استخراج یک بازنمایی نمادی برلی توصیف و مدلسازی یک سیستم معرفی شد.این نگاشت ها حوزه های خطی را با استفاده از گره های مفهوم ها (متغیرها، حالت ها) و رابطه های علامت دار فازی بین آنها توصیف می کنند.
کلمات کلیدی:
Application of evolutionary fuzzy cognitive maps to the long-term prediction of prostate cancer Wojciech Froelicha,1, Elpiniki I. Papageorgioub,∗, Michael Samarinas c, Konstantinos Skriapas c a Department of Informatics and Material Science, Institute of Computer Science, University of Silesia, ul.Bedzinska 39, Sosnowiec, Poland b Department of Informatics and Computer Technology, Technological Educational Institute of LAMIA, 3rd Km Old National Road Lamia-Athens, Lamia 35100, Greece c Department of Urology, General Hospital of Larisa, Larisa, Greece a r t i c l e i n f o Article history: Received 9 July 2011 Received in revised form 10 November 2011 Accepted 8 February 2012 Available online xxx Keywords: Prediction Prostate cancer Fuzzy cognitive maps a b s t r a c t The prediction of multivariate time series is one of the targeted applications of evolutionary fuzzy cognitive maps (FCM). The objective of the research presented in this paper was to construct the FCM model of prostate cancer using real clinical data and then to apply this model to the prediction of patient’s health state. Due to the requirements of the problem state, an improved evolutionary approach for learning of FCM model was proposed. The focus point of the new method was to improve the effectiveness of longterm prediction. The evolutionary approach was verified experimentally using real clinical data acquired during a period of two years. A preliminary pilot-evaluation study with 40 men patient cases suffering with prostate cancer was accomplished. The in-sample and out-of-sample prediction errors were calculated and their decreased values showed the justification of the proposed approach for the cases of long-term prediction. The obtained results were approved by physicians emerging the functionality of the proposed methodology in medical decision making. © 2012 Elsevier B.V. All rights reserved. 1. Introduction There are many knowledge representation methods known as connectionist methods [1]. From the point of view of their relationships to source data at least two approaches to their construction can be distinguished. The first type of networks possesses input and output nodes that represent data acquisition places and control points within problem environment respectively. As an example of such type of networks, we would like to mention artificial neural networks (ANN). They consist of input, output and hidden nodes (neurons). The main task of ANN is the approximation of function between input and output nodes. The ANNs represent a black-box function between input and output nodes, the relationships between nodes do not follow any interpretation issue. The second type of networks could be called as conceptual structures [2]. The intention of constructing conceptual structures is the representation of relationships between concepts. The nodes that This document is a collaborative effort. ∗ Corresponding author. E-mail addresses: wojciech.froelich@us.edu.pl, froelich@konto.pl (W. Froelich), epapageorgiou@teilam.gr (E.I. Papageorgiou), mikesamih@hotmail.com (M. Samarinas), kostas.skriapas@hotmail.com (K. Skriapas). 1 Principal corresponding author. represent concepts and the arcs that represent relationships are able to follow semantical interpretetion. The construction of conceptual networks could be accomplished on diverse levels of data abstraction. For example, the ontologies are built on symbolic level and are easily interpreted by humans. One of the directions of research on conceptual structures is the modeling of cause-and-effect relationship. In spite of years of intuitive and formal analysis, modeling of causality is still raising the interest of researchers. The main motivation is the expectation that the causal relationship, hidden in data reflect some stable mechanisms that can be discovered and applied for making predictions. Recently, the representation of causal relationships in the form of FCM is among the most active directions of research [3–6]. Due to their simplicity, supporting of inconsistent knowledge, and circle causalities for knowledge modeling and inferring, FCMs have found large applicability to many diverse scientific areas [7,6,8,5]. The works of Stach et al. [12] and Song et al. [21] addressed the problem of multivariate time series prediction. In this paper, our interest is focused on a conceptual structure and soft computing methodology which is FCM as proposed by [9]. FCM is represented by a graph with nodes representing concepts and directed arcs representing causal relationships between nodes. The FCM model exposes some similarities to ANNs. However, the main difference is the semantics. The FCM model is transparent in such meaning, that every node and edge within the FCM graph can be interpreted by a human. The FCM represents common-sense
