Please use this identifier to cite or link to this item:
http://dspace.dtu.ac.in:8080/jspui/handle/repository/19135
Title: | DIMENSION REDUCTION FOR SPAM SMS CLASSIFICATION |
Authors: | GUPTA, SHWETA |
Keywords: | DIMENSION REDUCTION SMS CLASSIFICATION MOBILE PHONES FEATURE SELECTION ALGORITHMS SPAM |
Issue Date: | May-2022 |
Series/Report no.: | TD-5722; |
Abstract: | The invent of technology has given us a lot of outcomes. With the availability of network or internet a lot of change had been brought to the life of people. Earlier we were all depending on the physical modes of communication, commodities and interactions. But nowadays, we are all equipped with online or virtual world. We have online friends, online transactions, markets, finances and everything. Social Media and Mobile devices are one of the after-effect that were introduced with all these emerging technologies. A lot of data is generated every second from various online platforms. These data are highly valuable for business purpose to numerous organizations that are continuously studying this data. SMS( Short Service Message) it is a text communication system where a person use to share information, data, news and communicate with his friend, family or other professional chats. As we know that with there is huge growth in the demand of mobile devices almost every person either from different financial background, of different age groups, belonging to different social communities, associated with different geographical areas or indulge in different economic activities owns these gadgets. Most of the population is owing Mobile phones nowadays. It had been found that the SMS or native messaging platform is one of the most preferred way of communication by different groups of people or organizations. With this the person can easily communicate and acquire knowledge he desires. But with some positive sides it had been ruined with some malevolent people or organization who tries to spoil the network by sending the rogue or undesired data to the recepients. These unsolicited SMS messages are termed as SPAM. As we are aware that the mobile phones are having limited memory and filling that with such malicious data is not something we want. Also, these messages that are received in bulk to the person may annoy him/her and may lead to skip of any urgent message that needs some earliest response or actions. A lot of Spam filtering tools are developed by numerous organizations but still the spammers had found some way to break the security. The spammers used to send the bulk of messages that tries to fill the user’s inbox, some of the messages include promotional advertisements, some include fake offers that ask the user to reveal his personal information that would be used by the them to provide some financial rewards or some interesting vouchers, some may include the malicious links that may lead to the sites or software that would steal all private data of the user to the spammer by which he can make the financial loss to the person or the organization or some critical information loss. The spammers earns a good amout of financials through these activities. So we need to eliminate all such unwanted data from the network. Here, in this work we had proposed an ensemble feature selection algorithm for the SPAM SMS classification that would help in classification of all such unsolicited messages from the platform. As there are large number of messages that are present in the database, we instead to manually doing such elimination move towards some v automated ways. We usually employ machine learning algorithms for the task. But due to the presence of large number of features the machine learning models may sometime be guided by mischievous features. The large number of features may also lead to the high computation cost. So, to remove such redundant or irrelevant features from the dataset we employ feature selection algorithm. The Proposed feature selection algorithm’s performance had been tested against 5 chosen feature selection algorithms. Also the obtained dataset from all the feature selection algorithm is fed to to the three machine learning models. The results obtained were compared with the one deep learning model too. |
URI: | http://dspace.dtu.ac.in:8080/jspui/handle/repository/19135 |
Appears in Collections: | M.E./M.Tech. Computer Engineering |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
SHWETA GUPTA M.TEch..pdf | 2.04 MB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.