توضیحات
چکیده
مقدمه
امروزه با پیشرفت تکنولوژی، برنامه های کاربردی حجم زیادی از جریان داده را با سرعت زیاد تولید میکنند. به عنوان مثال ترافیک شبکه، جریان کلیکهای وب، ویدئوهای نظارتی و شبکه های سنسور. کاوش جریان داده یک زمینه تحقیقاتی جدید محسوب میشود که هدف آن استخراج الگوها/ دانش پنهان از جریان داده پیوسته است. برخلاف روشهای دادهکاوی کلاسیک که در آنها مجموعه داده ها ایستا میباشند و میتوانند بارها مورد دسترسی قرار گیرند، الگوریتم های جریان کاوی با چالش های زیادی مانند محدودیت حافظه، پاسخ دهی بلادرنگ ،پویش تک گذره و تشخیص تغییر مفهوم روبرو هستند.
ABSTRACT
The flow of data to data that is said to be generated continuously and from real-time systems are collected and analyzed. The k-means algorithm is a common clustering method for analyzing data flow. The algorithm is simple, scalable and successful in real-time applications, but along with these advantages, it has a major limitation, namely the determination of the number of clusters, k. In many cases, it is always difficult or even impossible to determine the value of k at the very beginning. In terms of optimization, clustering is a hard-NP grouping. Evolutionary algorithms are meta-bureaucratic methods that can provide optimal solutions to these issues. In this study, we have improved the rapid evolutionary algorithm for data flow clustering. In the proposed method, we use conflict-based learning to produce high-quality primary populations. We have also used a different method for the phase of mutation and selection. Comparison of the proposed algorithm and the STREAM-FEAC algorithm, based on the simplified benchmark, shows that the proposed algorithm performs significantly better.
INTRODUCTION
Today, with the advent of technology, applications generate a large amount of data flow at high speed. For example, network traffic, web clicks, monitor videos and sensor networks. Exploring the data flow is a new research field aimed at extracting hidden patterns / knowledge from the continuous data stream. Unlike classical data mining methods in which datasets are static and can be accessed frequently, flow-generation algorithms face many challenges, such as memory limitation, real-time response, single-scan, and change-definition detection.
Year: 2018
Publisher : Third Annual National Conference on Electrical, Computer and Bioelectric Engineering in Iran
By : Marzieh Damour, Adel Ghazi Khani
File Information: persian Language/ 8 Page / size: 483 KB
Only site members can download free of charge after registering and adding to the cart
سال :1397
ناشر : سومین کنفرانس سالانه ی ملی مهندسی برق ,کامپیوتر و بیوالکتریک ایران
کاری از : مرضیه دامور ،عادل قاضی خانی
اطلاعات فایل : زبان فارسی / 8 صفحه / حجم : KB 483
نقد و بررسی‌ها
هنوز بررسی‌ای ثبت نشده است.