International Core Journal of Engineering 2020-26 | Page 126

2019 International Conference on Artificial Intelligence and Advanced Manufacturing (AIAM) CIMR: Intelligent Cellphones Recommendation for Consumers Based on Machine Learning Techniques Ruxun Xiang 1 , Min Fu 2, 3 1 Chengdu University of Information Technology, Chengdu, China 2 Alibaba Group, Hangzhou, China 3 School of Computing, Macquarie University, Sydney, Australia [email protected], [email protected] questions related to the following aspects: 1) informants’ identity and personal information, such as their ages, and the academic degree; 2) configurations of the phone they are using or they have used, such as operating system, and camera; 3) satisfaction degree about their present cellphone and their ideal cellphones. Furthermore, CIMR is based on techniques of machine learning. Abstract—Cellphones are electronic devices that people use most in daily life. There are various kinds of cellphones in the market nowadays, while ordinary consumers are dazzled by the huge cellphone market and do not know how to choose a suitable one. Hence, it is of great importance to help them in doing this kind of selection to satisfy their needs. This paper proposes a machine learning based cellphones recommendation approach, CIMR, for consumers to pick up the most suitable cellphones that can mostly meet their desired requirements. We evaluate CIMR by using thousands of real consumers all over the world with two proposed methods: Gradient Boosting Decision Tree (GBDT) and Random Forest (RF). The experimental results show that the accuracy rate for recommending cellphones by using our method is 95.87%, which exceeds the existing old methods by up to 5.49%. First, we analyze tens of thousands of realistic surveyed data related to consumers, cellphone configurations and cellphone usage. Second, we apply machine learning methods to train the surveyed data and obtain a model that represents the relationship between consumers and cellphones. Next, we apply a Random Forest algorithm and a Gradient Boosting Decision Tree algorithm to figure out the most suitable cellphones for each consumer that satisfy the requirements of him/her [5][6]. According to our knowledge, it is one of the first trials that machine learning techniques are applied on cellphone recommendation. Keywords—Machine learning, Cellphone Recommendation, Data Mining I. I NTRODUCTION We evaluate the feasibility and validity of CIMR through real-world scenarios-based experiments. In the experiments, we evaluate four sets of recommendation approaches using our real-world dataset, and these four approaches are: 1) two existing old methods: Naïve Bayes and Neural Network [15] [16]; 2) our newly proposed approach CIMR: Random Forest and Gradient Boosting Decision Tree. We make a comparison between the four approaches. Next, we select the better one of our methods and the better one of the existing methods to compare the cellphone recommendation accuracy results. The experimental results show that CIMR can increase the accuracy rate for recommending cellphones to 95.87%, which exceeds the existing old method by 5.49%. Advanced cellphone technologies have been changing rapidly and a large number of people purchase cellphones to meet their daily requirements, which leads to the popularity of cellphones [1]. The PC market was exceeded by the cellphone market in 2011 for the first time [1]. A related survey shows that the cellphones sold in the third quarter of 2013 increased by 44% [1]. Large-scale worldwide mobile industry giants, such as Apple and Huawei, are introducing new cellphones constantly with original functionalities and hardware facilities to attract more customers [2][3][4]. The common ways for a consumer to choose a cellphone are usually by watching commercials or asking friends and shopping guides for suggestions. There are several drawbacks: 1) Commercials exaggerate the descriptions of cellphones; 2) People’s friends do not understand their true demands; 3) Shopping guides make recommendations according to their own interests. There lacks a direct method to determine the accurate conditions and demands for each consumer. The contributions of this paper are: 1) We propose an original and intelligent machine learning based approach for recommending appropriate cellphone for consumers; 2) We propose a methodology to build up a connection between consumers’ personal situations and cellphones; 3) We propose a systematic evaluation way in the experiment to demonstrate how to evaluate the validity of our method. In this paper, we propose a novel and intelligent approach, named CIMR, for the following motivations: 1) recommending suitable cellphones to consumers for a maximization in budget use and an enjoyable purchase experience; 2) sending feedbacks of customers to manufacturers to make adjustments in cellphone configurations and sales strategies. CIMR is based on analysis of online and offline data we surveyed during the period from Sept. to Nov. of 2018. The survey contains 978-1-7281-4691-1/19/$31.00 ©2019 IEEE DOI 10.1109/AIAM48774.2019.00028 The remainder of this paper is: Section II introduces the background; Section III discusses the related work; Section IV illustrates our proposed method of recommending cellphones; Section V presents the experimental evaluation; Section VI provides the conclusion and our future work. We 104 introduce II. B ACKGROUND the importance of a cellphone