短信服務(wù)(SMS)在現(xiàn)代通信技術(shù)中得到了廣泛的推廣。短消息服務(wù)組件是現(xiàn)代社會(huì)中最快、最常用的電子消息發(fā)送方法。垃圾郵件或未經(jīng)請(qǐng)求的短信已成為組織、網(wǎng)絡(luò)系統(tǒng)和私人客戶(hù)端的一個(gè)值得注意的問(wèn)題。通過(guò)垃圾短信,垃圾郵件發(fā)送者正在影響時(shí)間和內(nèi)存空間,這是計(jì)算世界中最重要的資產(chǎn)。垃圾郵件的分類(lèi)是一個(gè)有趣而突出的問(wèn)題。這里介紹了與垃圾郵件相關(guān)的問(wèn)題以及努力管理垃圾郵件的不同方法。對(duì)SMS中的垃圾郵件可用性進(jìn)行分類(lèi)是一項(xiàng)具有挑戰(zhàn)性的任務(wù),因此,在這方面已經(jīng)進(jìn)行了大量的研究,這些研究采用了機(jī)器學(xué)習(xí)技術(shù),如樸素Bayes(NB)、隨機(jī)森林(RF)和支持向量機(jī)(SVM),用于垃圾郵件分類(lèi)。雖然這些方法表現(xiàn)出了足夠的性能,但在垃圾郵件分類(lèi)方面效率不夠。因此,需要進(jìn)行嚴(yán)格的研究,以找到更準(zhǔn)確、更穩(wěn)健的方法。為了解決這個(gè)問(wèn)題,我們提出了一種新的長(zhǎng)期短期記憶(LSTM)方法,它是一種具有包括記憶細(xì)胞在內(nèi)的門(mén)控機(jī)制的遞歸神經(jīng)網(wǎng)絡(luò)(RNN)的高級(jí)結(jié)構(gòu)。此外,本研究還采用了Word2Vec工具,該工具將簡(jiǎn)化文本轉(zhuǎn)換為向量空間中單詞的表示形式。為了評(píng)估我們的方法的有效性,SMS數(shù)據(jù)集已被免費(fèi)使用。實(shí)驗(yàn)結(jié)果表明,該方法優(yōu)于最...

【文章頁(yè)數(shù)】：52 頁(yè)

【學(xué)位級(jí)別】：碩士

【文章目錄】：
摘要
Abstract
1 Introduction
    1.1 What is Spam
    1.2 Spamming Motivations
    1.3 Research Background and Significance
        1.3.1 Research Background
        1.3.2 Research Significance
    1.4 Overseas and Domestic Research Progress
        1.4.1 Spam Detection in Short Message Service(SMS)
        1.4.2 Spam Detection in Email
    1.5 Main Contents and Structure of the Thesis
        1.5.1 Main Contents
        1.5.2 The Structure of the Thesis
    1.6 Summary
2 Basic Theory and Related Work
    2.1 Basic Theory of Machine Learning
        2.1.1 Unsupervised Learning
        2.1.2 Supervised Learning
    2.2 Spam Filtering Techniques
        2.2.1 Machine Learning Approach to Spam Filtering
        2.2.2 Artificial Neural Network
        2.2.3 Deep Neural Network
    2.3 Spam Filtering Challenges for Machine Learning
        2.3.1 False Positive
        2.3.2 Concept Drift Handling in SMS
        2.3.3 E-mail Ranking or Prioritizing
    2.4 Summary
3 Experimental Model
    3.1 Proposed method LSTMs
    3.2 Word Embedding
    3.3 Word2Vec
        3.3.1 Skip-gram Model
        3.3.2 Continuous Bag-of-Words(CBOW)Model
    3.4 Data Set
    3.5 Traditional Baseline Methods
        3.5.1 SVM(Support Vector Machine)
        3.5.2 Decision Tree
        3.5.3 KNN(K-Nearest Neighbors)
        3.5.4 Random Forest
        3.5.5 NB(Na?ve Bayes)
    3.6 Summary
4 Results and Discussions
    4.1 Spam Detection Framework
        4.1.1 Detecting Strategy
        4.1.2 Contributions
        4.1.3 Data Preprocessing
    4.2 Comparative Study of Results
        4.2.1 Comparative Results
        4.2.2 Detecting Results
5 Conclusion and Future Work
References
Research Projects and Publications in Master Study
Acknowledgement



本文編號(hào)：3801134