International Core Journal of Engineering 2020-26 | Page 130

D. Discussion
We next provide discussions for the possible reasons for
our outstanding performances.
Our future work includes: 1) introduce more features to
better describe the personal information and cellphone
characteristics more completely and precisely; 2) connect
CIMR with front-end browsers or cellphone applications to
realize an actual use in market; 3) try applying CIMR on
other recommendation activities and conduct research on
adjusting to cater those activities. 4) obtain more data as
training data, for a more accurate prediction model.
First, for the Gradient Boosting Decision Tree, it takes
the Decision Tree as the weak classifier in Gradient
Boosting Modeling (GBM) [20]. Hence, GBDT benefits
from good characteristics of Decision Tree itself [20].
GBDT can deal with various types of features, and with
little disturbance if some irrelevant features exist in the
dataset. Consequently, it can both deal with continuous and
discrete values, which suits our preprocessed dataset
perfectly. Finally using some robust cost functions, GBDT
has strong robustness on abnormal values, or, it is very
inclusive[21]. Therefore, the GBDT method suits the given
discrete dataset with different types of features.
R EFERENCES
[1]
[2]
[3]
For Random Forest, it can deal with a dataset of many
features without doing feature selection [19]. Besides, RF
uses the method of bagging algorithm, integrating a variety
of non-dependent weak learners into a strong learner. Also,
Sampling is used every time in the training, so it owns a
strong generalization ability which is useful for reducing the
variance of the model. And with the use of Majority Voting,
RF combines the prediction results of multiple weak
learners and select the best results to avoid the predicting
bias. Thus, RF does an excellent job with such a large
feature scale and unbalanced dataset originally.
[4]
[5]
[6]
From the cellphone recommendation results of our
proposed method in Table I, we obtained a highly effective
result which satisfies the purchase needs for most consumers
with the accuracy of 95.87% and exceeds 5.49% compared
with the existing old method. Moreover, we noticed from
Fig. 4 and Fig. 5 that, within certain errors, statistical
distributions for the predicted data and the original data are
highly similar in shape. Some differences exist because it is
hard to include all types for a cellphone model and some
outdated cellphones are not included in the features. [7]
To sum up, our proposed method shows an accurate
prediction and a highly resembled data distribution, which
confirms our valid and effective method again. Hence,
CIMR can satisfy the recommendation requirements for
purchasing cellphones. [11]
[8]
[9]
[10]
[12]
[13]
VI. C ONCLUSION AND F UTURE W ORK
[14]
Varieties of recommending system compete for the
market fiercely, such as Mobile application recommendation,
shopping bundle recommendation. However, we notice that
there are vacancies in the cellphone recommendation this
paper proposes a cellphone intelligent recommending
framework, called CIMR, for the general public to make
satisfactory cellphones purchasing. We design and
implement CIMR based on machine-learning-based
techniques, and online and offline surveyed data. It is
proved to be able to make appropriate recommendations on
cellphones. We evaluate CIMR in comparison with actual
cellphone usage. The evaluation results show that our
method can increase the cellphone recommendation
matching degree by 5.49% and to a large extent construct
the distribution of predicted cellphones in accordance with
the actual cellphones distribution.
[15]
[16]
[17]
[18]
[19]
[20]
[21]
108
B. Liu, et al. 2016. Structural Analysis of User Choices for Mobile
App Recommendation. ACM Transactions on Knowledge Discovery
from Data, Vol.11, No.2, Article 17. ACM, New York, NY, USA.
http://dx.doi.org/10.1145/2983533.
Apple
Discussion
Forum.
https://discussionschinese.apple.com/browse.
H. Zhang, H.G. Rao and J.Z. Feng. 2018.Product innovation based on
online review data mining: a case study of Huawei phones. Electronic
Commerce Research Vol.18 Iss.7, 19 pages Springer, US.
P. Samuelson. Apple v. 2016. Samsung and the upcoming design
patent wars? Communications of the ACM, Vol.59 Iss.7.ACM, New
York, NY, USA, 3 pages.
F. J. R. Ramírez, et al. 2018. Assessment of Antarctic moss health
from multi-sensor UAS imagery with Random Forest Modelling.
International Journal of Applied Earth Observation and
Geoinformation, Vol.68, 10 pages.
J. Y. Chen, et al. 2018. Predicting Drug Target Interactions Based on
GBDT. Machine Learning and Data Mining in Pattern Recognition,
10pages. Springer, Cham.
B. S. Kumar and M. Ryoke. 2013. Analyzing the Business Strategies
of Mobile Phone Operators Using Agent-Based Simulation.
International Journal of Knowledge and Systems Science. 8 pages.
K. Hui, et al. 2018. Personalized trip recommendation for tourists
based on user interests, points of interest visit durations and visit
recency. Knowledge and Information Systems. Springer, 12 pages.
E. Kim and M. Kim. 2016. Topic-tracking-based dynamic user
modeling with TV recommendation applications. Applied Intelligence.
Springer, US, 21 pages.
S. Kim, B. Min Mun and Suk Joo Bae. 2018. Data depth based
support vector machines for predicting corporate bankruptcy. Applied
Intelligence. Springer, US, 13 pages..
M. Fu, et al. 2018. DAliM: Machine Learning Based Intelligent
Lucky Money Determination for Large-Scale E-Commerce
Businesses. The 16th International Conference on Service-Oriented
Computing (ICSOC 2018), Hangzhou, China,15 pages.
Xiao Mi Forum.2019. http://www.xiaomi.cn/.
China
Smartphone
Market
Research
Report.2018.
http://www.iimedia.cn/60944.html.
Mobile phone camera module market research report.
http://m.elecfans.com/article/708419.html.
N. S. Harzevili and S.H. Alizadeh. 2018. Mixture of latent
multinomial naive Bayes classifier. Applied Soft Computing. Vol.69.
11 pages 516-527. https://doi.org/10.1016/j.asoc.2018.04.020.
A.L. Caterini and D.E. Chang. 2018. Deep Neural Networks in a
Mathematical
Framework.
Springer,
Cham.
https://doi.org/10.1007/978-3-319-75304-1.
Aliyun
official
developer
technology
community.
https://blog.csdn.net/yunqiinsight/article/details/80108816.
China
Software
Development
Network.
https://blog.csdn.net/panjiao119/article/details/80311529?tdsourcetag
=s_pctim_aiomsg.
China
Software
Development
Network.
https://blog.csdn.net/zhongjunlang/article/details/79488955.
China
Software
Forum.
https://www.cnblogs.com/bentuwuying/p/6667267.html.
China
Software
Forum.
https://www.cnblogs.com/pinard/p/6140514.html