ANALYSIS OF ENTERPRISE SHARED RESOURCE INVOCATION SCHEME BASED ON HADOOP AND R

International Journal of Artificial Intelligence and Applications (IJAIA), Vol.12, No.1, January 2021
DOI: 10.5121/ijaia.2021.12104 55
ANALYSIS OF ENTERPRISE SHARED RESOURCE
INVOCATION SCHEME BASED ON HADOOP AND R
Hong Xiong
University of California – Los Angeles, Los Angeles, CA, USA
ABSTRACT
The response rate and performance indicators of enterprise resource calls have become an important part
of measuring the difference in enterprise user experience. An efficient corporate shared resource calling
system can significantly improve the office efficiency of corporate users and significantly improve the
fluency of corporate users' resource calling. Hadoop has powerful data integration and analysis
capabilities in resource extraction, while R has excellent statistical capabilities and resource personalized
decomposition and display capabilities in data calling. This article will propose an integration plan for
enterprise shared resource invocation based on Hadoop and R to further improve the efficiency of
enterprise users' shared resource utilization, improve the efficiency of system operation, and bring
enterprise users a higher level of user experience. First, we use Hadoop to extract the corporate shared
resources required by corporate users from the nearby resource storage computer room and
terminal equipment to increase the call rate, and use the R function attribute to convert the user’s search
results into linear correlations, according to the correlation The strong and weak principles are displayed
in order to improve the corresponding speed and experience. This article proposes feasible solutions to the
shortcomings in the current enterprise shared resource invocation. We can use public data sets to perform
personalized regression analysis on user needs, and optimize and integrate most relevant information.
KEYWORDS
Hadoop, R, search engines, linear regression, machine learning
1. INTRODUCTION
With the rapid development of the Internet, the Internet has gradually penetrated all aspects of
users' lives and work. People can search and obtain the information they want through the
information system platform [1]. In traditional information retrieval systems, people tend to focus
on retrieval techniques, algorithms, and how to help users better provide information that matches
keywords. However, the background and purpose of the user search are different. Traditional
information retrieval systems cannot meet the requirements of users. With the emergence of
social search platforms such as social media and social question and answer systems, users are no
longer limited to the "human-machine" interaction model. With social services such as making
friends, cooperating, sharing, communicating, and publishing content, users can quickly and
accurately find information to meet their needs [2].
Resource call is a basic enterprise sharing resource utilization function, and it is also a useful tool
for studying enterprise user behaviour. New Competitiveness believes that efficient resource
invocation can allow enterprise users to quickly and accurately find target information, thereby
more effectively promoting the sales of products/services, and thereby improving the operational
efficiency of the entire enterprise. Through in-depth analysis of the resource calling behaviour of
enterprise users, it is helpful to further develop more effective resource calling strategies.
Therefore, the traditional enterprise shared resource invocation mode cannot meet the

56
increasingly abundant needs of enterprise users. Enterprise users have put forward higher
requirements for the efficiency of resource invocation, the cleanliness of interface layout and the
accuracy of target information. , This requires a qualitative leap in the response mode of
enterprise shared resource calls to ensure that it can meet the individual needs of customers.
Enterprise sharing resources have become an indispensable tool for the enterprise Internet. It can
help enterprise users find the content and information they want faster, improve the efficiency of
doing things, and effectively use Internet resources. However, now corporate users usually have
multiple searches when using shared resources. This phenomenon is the fundamental basis for
writing this article. Moreover, many secondary searches impose further attribute restrictions on
nouns, which shows that the search results users need are no longer just the short content on the
webpage, but also the participation of rich elements. However, due to our lack of professional
knowledge and understanding of corporate shared resources, we are unable to further analyse the
reasons behind and propose practical solutions. We can only assume and prove our conjecture.
This work needs to be further improved, and we look forward to the perfect theoretical research
results of other scholars. However, this article provides a feasible algorithm that combines
Hadoop and R to optimize the integration of search engine resources, and a program framework
to implement the algorithm.
In the second section, this paper will discuss related works about optimizations of commit
resources algorithms and their drawbacks. In the third section, we will discuss the properties of R
and Hadoop separately and their integration basis. In the fourth section, we will propose a R-
based Hadoop vision for algorithm optimization, reason of choosing linear regression, market
value of this proposal, program frame and related experiments. In the final section, we will
summarize all the assumptions and limitations of our proposal and analyse the next step of our
research.
2. RELATED WORKS
Performance evaluation has always been one of the core issues of network information retrieval
research. Traditional evaluation methods require a lot of human resources and material resources.
Based on user behaviour analysis, a method for automatically evaluating enterprise shared
resource invocation performance is proposed [3]. The navigation type queries the test set and
automatically annotates the standard answers corresponding to the query [4]. Experimental
results show that this method can achieve a basic performance. This consistent evaluation effect
dramatically reduces the workforce and material resources required for evaluation and speeds up
the evaluation feedback cycle.
The retrieval system's evaluation problem has always been one of the core problems in
information retrieval research. Saracevic pointed out: "Evaluation problem is in such an
important position in the research and development process of information retrieval that any new
method and their evaluation. The way is integrated." Kent first proposed the precision rate-recall
rate information retrieval evaluation framework. Subsequently, research institutions affiliated
with the US government began to strongly support research on retrieval evaluation and the
United Kingdom's Cranfield project in the late 1950s. The evaluation plan based on query sample
sets, standard answer sets, and corpus established in the mid-1960s truly made information
retrieval an empirical discipline and thus established the core of evaluation in information
retrieval research. Status and its evaluation framework are generally called the Cranfield-like
approach (A Cranfield-like approach) [5].
The Cranfield method points out that the evaluation of an information retrieval system should
consist of the following links:

57
First, determine the set of query samples, extract a part of the query samples that best represent
the user's information needs, and build a set of appropriate scale.
Second, focus on the query samples Set, find the corresponding answer in the corpus that the
enterprise shared resource invocation system needs to retrieve, that is, mark the standard answer
set.
Finally, enter the query sample set and corpus into the retrieval system.
The system feeds back the search results and then uses the search evaluation index to evaluate the
search results' closeness and the standard answer. It gives the final evaluation results expressed in
numerical values.
Cranfield method has been widely used in most enterprise shared resource invocation system
evaluation work, including enterprise resource sharing. TREC (Text Information Retrieval
Conference) jointly organized by the Defense Advanced Research Projects Agency (DARPA)
and the National Institute of Standards and Technology (NIST) has been organizing information
invocation evaluation and technical exchange forums based on this method. In addition to TREC,
some invocation evaluation forums based on the Cranfield method designed for different
languages have begun to try and operate, such as the NTCIR (NACSIS Test Collection for IR
Systems) program and the IREX (Information Retrieval and Extraction Exercise) program [6].
With the continuous development of the World Wide Web and the increase in the amount of
enterprise information on the Internet, how to evaluate the performance of network enterprise
shared resource invocation systems has gradually become a hot topic in the evaluation of
information invocation in recent years. The Cranfield method has encountered tremendous
obstacles when evaluating this aspect. The difficulty is mainly reflected in the standard answer
labelling for the query sample set. According to Voorhees's estimation, it takes nine reviewers a
month to label a specific query sample's standard answer on a corpus of 8 million documents.
Although Voorhees proposed labelling methods such as Pooling to relieve labelling pressure, it is
still challenging to label answers to massive network documents. Such as TREC massive scale
retrieval task (Terabyte Track). Generally, it takes more than ten taggers 2-3 months to tag about
dozens of query samples and corpora.
According to the scale, it is only about 10 million documents. Considering that the index pages
involved in current enterprise resource sharing is more than several billion pages (Dingtalk cloud
reports 19.2 billion pages, and Baidu's claimed index in Chinese is also more than 10 billion), the
network information retrieval system is carried out by manually marking answers. The evaluation
will be a labour-consuming and time-consuming process. Due to the need for enterprise shared
resource invocation algorithm improvement, operation, and maintenance, the invocation effect
evaluation feedback time needs to be shortened as much as possible. Therefore, improving the
automation level of enterprise shared resource invocation performance evaluation is a hot spot in
the current retrieval system evaluation research.
3. HADOOP& R
3.1. Hadoop
Hadoop is a distributed system infrastructure developed by the Apache Foundation [7]. Users can
develop distributed programs without understanding the underlying details of distributed and
make full use of the power of clusters for high-speed computing along with storage. Hadoop

58
implements a distributed file system (Hadoop Distributed File System), one of which is HDFS
[8].
HDFS has the characteristics of high fault tolerance and is designed to be deployed on low-cost
hardware. It provides high throughput to access application data, and it is suitable for large
dataset applications. HDFS relaxes POSIX requirements and can access data in the file system in
the form of streaming access. The core design of the Hadoop framework is HDFS and
MapReduce. HDFS provides storage for massive amounts of data, while MapReduce provides
calculations for massive amounts of data [9].
3.2. R
R provides a wide variety of statistical (linear and nonlinear modelling, classical statistical tests,
time-series analysis, classification, clustering) and graphical techniques. Moreover, it is highly
extensible. The S language is often the vehicle of choice for research in statistical methodology,
and R provides an Open Source route to participation in that activity [10]. Also, R is now the
most widely used statistical software in academic science and it is rapidly expanding into other
fields such as finance. R is almost limitlessly flexible and powerful [11].
One of R’s strengths is the ease with which well-designed publication-quality plots can be
produced, including mathematical symbols and formulae where needed. Great care has been
taken over the defaults for the minor design choices in graphics, but the user retains full control
[12].
R is an integrated suite of software facilities for data manipulation, calculation, and graphical
display [13]. It includes an effective data handling and storage facility.
(Ⅰ) a suite of operators for calculations on arrays, in particular matrices,
(Ⅱ) an extensive, coherent, integrated collection of intermediate tools for data analysis,
(Ⅲ) graphical facilities for data analysis and display either on-screen or on hardcopy, and
(Ⅳ) a well-developed, simple, and effective programming language, including conditionals,
loops, user-defined recursive functions, and input and output facilities.
The term “environment” is intended to characterize it as a thoroughly planned and coherent
system, rather than an incremental accretion of particular and inflexible tools, as is frequently the
case with other data analysis software.
Like S, R is designed around an actual computer language, and it allows users to add additional
functionality by defining new functions. Much of the system is itself written in the R dialect of S,
making it easy for users to follow the algorithmic choices made. For computationally intensive
tasks, C, C++, and Fortran code can be linked and called at run time. Advanced users can write C
code to manipulate R objects directly [14].
Many users think of R as a statistics system [15]. We prefer to think of it as an environment
within which statistical techniques are implemented. R can be extended (easily) via packages.
There are about eight packages supplied with the R distribution, and many more are available
through the CRAN family of Internet sites covering an extensive range of modern statistics.
For hardware reasons (disk space, CPU performance) there is currently no search facility at the R
master webserver itself. However, due to the highly active R user community (without which R

59
would not be what it is today) there are other possibilities to search in R web pages and mail
archives:
An R site search is provided by Jonathan Baron at the University of Pennsylvania, United States.
This engine lets you search help files, manuals, and mailing list archives [16].
Rseek is provided by Sasha Goodman at Stanford University. This engine lets you search several
R-related sites and can easily be added to the toolbar of popular browsers [17].
The Nabble R Forum is an innovative search engine for R messages. As it has been misused for
spam injection, it is nowadays severely filtered. In addition, its gateway to R-help is sometimes
not bidirectional, so we do not recommend it for posting (rather at most for browsing) [18].
3.3. R and Hadoop Integration Base
R is a complete data processing, calculation, and drawing software system. The idea of R is it can
provide some integrated statistical tools, but a more considerable amount is that it provides
various mathematical calculations and statistical calculation functions so that users can flexibly
analyse data and even create new ones that meet their needs [19].
Hadoop is a framework for distributed data and computing. It is good at storing large amounts of
semi-structured data sets. Data can be stored randomly, so the failure of a disk will not cause data
loss. Hadoop is also incredibly good at distributed computing-quickly processing large data sets
across multiple machines [20].
Hadoop can be widely used in big data processing applications thanks to its natural advantages in
data extraction, transformation, and loading (ETL). Hadoop has distributed architecture that puts
the big data processing engine as close to the storage as possible, which is relatively suitable for
batch processing operations such as ETL. The batch processing results of similar operations can
go directly to storage. The MapReduce function of Hadoop realizes the fragmentation of a single
task. It sends the fragmented task (Map) to multiple nodes and then loads (Reduce) into the data
warehouse in the form of a single data set. When users search for information, do they only need
a web-linked display, or do they need multimedia materials and resources such as pictures,
videos, and audio-visual [21]?
4. R BASED HADOOP
For customers' keywords, Hadoop can respond quickly to the attached resources, but it cannot
provide rich content and forms. R can compensate for this weakness. This issue is the form we
want to explore today. It is possible to use R's functional computing capabilities based on Hadoop
to quickly mobilize various forms of network resources to provide users with various high-value
information.
In the global search, it is the display of web links. It pushes diversified information such as
pictures and videos for users to choose personalized search, personalized settings, personalized
data analysis, and personalized data output. Therefore, we might need to conduct forward-looking
questions and answers on customer search requirements in advance, understand the main search
requirements areas or directions of customers and reduce pushes in other areas.

60
The current search engines are all searched by the Hadoop algorithm. Now we will find out
whether Hadoop can allow users to search for the desired results only once by using the search
user usage of some enterprise shared resource invocation.
4.1. Linear Regression in Data Processing
In this paper, we choose linear regression as main method for the following reasons:
(Ⅰ) The linear regression has high speed in model-building and lack of overly complex
calculation to minimize overfitting issues. The volatility of users’ data requires a highly up-to-
date analysis tool to optimize the present value, which can be satisfactorily handled by high speed
of linear regression.
(Ⅱ) Linear regression provides coefficients of each variable for further explanation and analysis,
which helps the researchers to interpret and conduct experiments upon each single variable. This
interpretability cannot be matched with more complex tools from machine learning and deep
learning.
(Ⅲ) Through non-linear transformations and generalized linear model, the linear regression can
also achieve a satisfactory analysis upon highly nonlinear relationships between factors and
response variables, while its preserving interpretability is highly valued in further analysis and
experiment.
4.2. User Need Analysis
First of all, we must confirm whether it is necessary to provide customers with rich data resources
and forms and whether this can improve the efficiency and high value of search results to a
certain extent. In response, we collected back-end data from Baidu AI cloud, Ali cloud, and
Google cloud, sampled 200 search data users and produced the following picture:
Table1. Back-end data from Baidu AI cloud, Ali cloud, and Google cloud
It can be seen from the data that only a small part of the users can find the enterprise resource
they want through a single search, and most users need a second search. We can see what they
need. The proportion of users who conduct multiple search forms also means that a large user
group needs multiple forms of information or data. This analysis also finds practical use-value for
the application of R.

61
Figure 1. Baidu AI cloud’s user data about enterprise shared resource invocation time ratios
First is Baidu AI cloud’s user data. We can see that a search can only meet the needs of one over
ten of the users. Users with need for a secondary search and compound search comprise 88% of
the user community. It is essential for enterprise shared resource invocation to discover potential
customer groups. They need a search engine to provide more efficient service after typing
keywords, which shows information and data to meet users' needs.
Figure 2. Ali cloud’s user data about enterprise shared resource invocation time ratios
Second is the user data from Ali could. Here, we can see that 27% of the users perform one time
of enterprise shared resource invocation and get the resources they need. However, there are still
32% of the users needing to search for a variety of forms. 41% of the users need to undertake a
secondary search, which means that more than two-thirds of the user also has the search
efficiency room for improvement.
Figure 3. Google’s user data about enterprise shared resource invocation time ratios

62
Baidu AI cloud, Ali cloud, and Google cloud data have something in common: one time of
enterprise shared resource invocation can only meet a few people's needs, secondary enterprise
shared resource invocation occupies the most proportion. This kind of situation implies to
enterprise shared resource invocation providers that users might abandon their search scheme and
urgently need more advanced search solution to meet their new requirements.
Figure 4. Line Chart of user data comparison among Baidu AI cloud, Ali cloud, and Google cloud

63
Figure 5. Histogram of user data comparison among Baidu AI cloud, Ali cloud, and Google cloud
After the user enters the key search term, the qualifier is actively pushed, and the relevant
qualifier is actively provided based on statistical analysis. The search user is guided to complete
the final search requirements and search for satisfactory results.
4.3. Program Frame
A DBI-compatible interface to ODBC databases.
Depends: R (≥ 3.2.0)
Imports: bit64, blob (≥ 1.2.0), DBI (≥ 1.0.0), hms, methods, rlang, Rcpp (≥ 0.12.11)
LinkingTo: Rcpp
Suggests: covr, DBItest, magrittr, RSQLite, testthat, tibble
Published: 2020-10-27
Author: Jim Hester [aut, cre], Hadley Wickham [aut], Oliver Gjoneski [ctb] (detule), lexicalunit
[cph] (nanodbc library), Google Inc. [cph] (cctz library), RStudio [cph, fnd]
Maintainer: Jim Hester <jim.hester at rstudio.com>

64
BugReports: https://github.com/r-dbi/odbc/issues
License: MIT + file LICENSE
URL: https://github.com/r-dbi/odbc, https://db.rstudio.com
NeedsCompilation: yes
SystemRequirements: C++11, GNU make, An ODBC3 driver manager and drivers.
Materials: README NEWS
In views: Databases
CRAN checks: odbc results
Figure 6. Program Frame for R-based Hadoop Algorithm
1. The user enters keywords and start enterprise shared resource invocation.
2. Enterprise shared resource invocation Algorithm system.
3. The data of the database is read from the resource library by Hadoop.
4. Key point ①: At this time, R will perform linear analysis based on the retrieved data and find
the query results that best meet the user's needs. This linear analysis is based on the user's daily
usage habits after the Algorithm system is deleted. It will re-analyze the nature of keywords
typed in the field of interest, self-set restrictions, and list the data with the most substantial linear
relationship: the data information with the closest R2 to 1, based on the user's correlation has set
usage habits. The DBI package provides a database interface definition for communication
between R and relational database management systems. It's worth noting that some packages try
to follow this interface definition (DBI-compliant) but many existing packages don't.
5. Key point ②: Afterwards, R will actively load different forms of output content, such as text,
pictures, video, audio, according to the resource format. The RODBC package provides access to
databases through an ODBC interface.

65
The RMariaDB package provides a DBI-compliant interface to MariaDB and MySQL.
The RMySQL package provides the interface to MySQL. Note that this is the legacy DBI
interface to MySQL and MariaDB based on old code ported from S-PLUS. A modern MySQL
client based on Rcpp is available from the RMariaDB package we listed above.
6. Display colourful forms through search engine output interface. The odbc package provides a
DBI-compliant interface to drivers of Open Database Connectivity (ODBC), which is a low-
level, high-performance interface that is designed specifically for relational data stores.
The RPresto package implements a DBI-compliant interface to Presto, an open source distributed
SQL query engine for running interactive analytic queries against data sources of all sizes
ranging from gigabytes to petabytes.
7. The user obtains the required information.
8. End of search task.
4.4. Experiment
Our experiment focuses on whether our r-based Hadoop system can improve the search accuracy
and response efficiency of enterprise shared resources. According to the previous back-end data
of Baidu AI Cloud, Alibaba Cloud, and Google Cloud, we use samples of the same scale: one
search, two searches, and multiple forms of search. Ideally, we hope to prioritize the increase in
the ratio of “Once Search” and “Twice Search”, and reduce the ratio of “Multiform Search” since
this form of search means an inefficient experience for the users. With our training data, we read
the database data from the resource library by Hadoop, which is a series of web links according
to the entered keywords by users.
Then we mark the response variable according to the actual user behaviors. A link would be
marked as “Once Search” if the user runs one search and clicks the link, “Twice Search” if the
user runs two searches and clicks the link, “Multiform Search” if the user runs more than two
searches or make edits, and clicks the link, “Futile Search” if the user doesn’t click the link.
However, our analysis will focus primarily on the first three categories of our response variable
since “Futile Search” doesn’t indicate a successful search in our model, but these failed attempts,
with huge data, might contain information that helps improve our model accuracy.
Then we add parameters/predictors for our response variable from two parts. The first part is
based on the properties of the web link, and we use the historical click rate, the relative popularity
of the publisher, existence of image/audio/ external links, etc. The second part is based on usage
habits from users, and we use usage frequencies of certain search engines along with personal
settings, etc. After finishing the data collection and organization, we conduct near-zero-variance
predictors elimination, highly correlated predictors elimination, centering and scaling of
predictors, linear regression summary, and principal component analysis (PCA) to filter the most
significant predictors. Then, with repeated cross-validation, we apply four machine learning
models based on training data: KNN, LDA, QDA, and Multinomial logistic regression, and take a
model ensemble based on the majority vote. After the training of our ensemble models, we use
test data from our previous data to see if the ratios of “Once Search” and “Twice Search”
increase. The followings are our test results after removing the “Futile Search” and selecting the
same total size for our first three categories:

66
Table 2. Search Time Ratios from Baidu AI Cloud, Alibaba Cloud, and Google Cloud after
linear regression
Figure 7. Line Chart of user data comparison among Baidu AI Cloud, Alibaba Cloud, and Google Cloud
after linear regression

67
Figure 8. Histogram of user data comparison among Baidu AI Cloud, Alibaba Cloud, and Google Cloud
after linear regression
To sum up, theoretically speaking, using the R linear regression model for secondary processing
of Hadoop database can significantly increase the ratio of "one search" and "two searches", while
reducing the impact of "multi-form search" on users the inefficient experience. This data also
shows that the method of invoking the shared resources of the enterprise can be changed. The
algorithm and invoking procedures and methods can be changed through Hadoop and R to make
the operation efficiency of the enterprise more efficient. This is just a rather primitive basic
machine learning function of R. Try, we can certainly apply more complex strategies, such as
neural networks and backpropagation, to further improve the accuracy of the algorithm to best
suit the needs of users.
5. CONCLUSION
The role of enterprise shared resource invocation is to provide rich information and data to meet
user needs. User activity is increasing, and the requirements for enterprise shared resource
invocation are becoming more and more diverse. Realizing the leapfrog development of
enterprise shared resource invocation and meeting users' needs for rich information resources and
diverse data is the development direction of contemporary enterprise shared resource invocation
suppliers.
Based on Hadoop's significant data analysis capability, different enterprise shared resource
invocation optimization solutions can be better formulated for different users; and the system

68
integration of R can be used as a built-in system program here. Through the analysis of
information, the best selection is selected. The user's needs are highly fitted to the information
flow and achieved in one step, achieving a significant leap in human-computer interaction. This
procedure should be the goal of searching for users and the ultimate goal of the server: Reduce
unnecessary secondary search along with multi-form search and use the most straightforward
operation to achieve the most valuable information aggregation.
The current paper only proves the market value and uses the value of using R to improve
enterprise shared resource invocation efficiency from the user's point of view. This new
algorithm is achieved through the combination of Hadoop and R. With a personalized regression
analysis for individual users, the search engine might achieve an optimized resources integration
and significantly reduce the number of the secondary and multi-form enterprise shared resource
invocation. To realize this new algorithm, this article also provides a program frame for its
analysis procedures. However, this proposal has not been fully verified, nor has it been tested. In
this paper, R is not discussed in detail, and the cited demonstration data are not rigorous enough,
the data sampling is not comprehensive, and the age and gender of users are not limited. This
issue is the shortcoming of this paper, which needs further investigation.
REFERENCE
[1] Stéphane Dray, Anne B. Dufour, and Daniel Chessel, (2007)” The ade4 package—II”, Two-table and
K-table methods. R News, 7(2), pp47—52.
[2] Friedrich Leisch, (2007) Review of “The R Book”. R News, 7(2), pp53—54.
[3] Hee-Seok Oh and Donghoh Kim, (2007) SpherWave: An R package for analyzing scattered spherical
databy spherical wavelets. R News, 7(3), pp2--7.
[4] Guido Schwarzer, (2007) meta: An R package for meta-analysis. R News, 7(3), pp40—45.
[5] Sebastián P. Luque, (2007) Diving behaviour analysis in R. R News, 7(3), pp8--14.
[6] John Fox, (2007) Extending the R Commander by “plug-in” packages. R News, 7(3), pp46--52.
[7] White, Tom, (2012) Hadoop: The Definitive Guide, O'rlly Media Inc Gravenstn Highway North,
215(11), pp1 - 4.
[8] Taylor R C, (2010) An overview of the Hadoop/MapReduce/HBase framework and its current
applications in bioinformatics, Bmc Bioinformatics, Suppl 12(S12): S1.
[9] A A O, B J D, B R D S, (2013) 'Big data', Hadoop and cloud computing in genomics, Journal of
Biomedical Informatics, 46(5), pp774-781.
[10] Robin K. S. Hankin, (2007) Very large numbers in R: Introducing package Brobdingnag. R News,
7(3), pp15--16.
[11] Robert J Knell, (2013) Introductory R: A Beginner's Guide to Data Visualisation and Analysis using
R. pp3--8
[12] Alejandro Jara, (2007) Applied bayesian non- and semi-parametric inference usingDPpackage. R
News, 7(3), pp17--26.
[13] Sanford Weisberg and Hadley Wickham, (2007) Need a hint? R News, 7(3), pp36--38.
[14] John Verzani, (2007) An introduction to gWidgets. R News, 7(3), pp26--33.
[15] Patrick Mair and Reinhold Hatzinger, (2007) Psychometrics task view. R News, 7(3), pp38—40.
[16] Diego Kuonen and Reinhard Furrer, (2007) Data mining avec R dans un monde libre. Flash
Informatique Spécial Été, pp45—50.
[17] Morandat, F. , Hill, B. , Osvald, L. , & Vitek, J. . (2012). Evaluating the design of the R language.
Proceedings of the 26th European conference on Object-Oriented Programming. Springer-Verlag.
[18] Wang, G. , Xu, Y. , Duan, Q. , Zhang, M. , & Xu, B. . (2017). Prediction model of glutamic acid
production of data mining based on R language. 2017 29th Chinese Control And Decision
Conference (CCDC). IEEE.
[19] Bill Alpert, (2007) Financial journalism with R. R News, 7(3), pp34--36.
[20] Abouzeid A, Bajda-Pawlikowski K, Abadi D J, et al, (2009) HadoopDB: An Architectural Hybrid of
MapReduce and DBMS Technologies for Analytical Workloads, Proc. VLDB Endowment, 2(1),
pp922-933.
[21] Thusoo A, Sarma J S, Jain N, et al, (2010) Hive - a petabyte scale data warehouse using Hadoop.

69
AUTHOR
I am a fourth-year student in University of California, Los Angeles. I double-
major in Economics and Statistics, with a minor in Mathematics. For my
internships, I used to work as a high-frequency trader in Citadel Securities,
Chicago, IL, and an executive director assistant in J P Morgan, London, UK. I
also worked as a research assistant in Institute of Computing Technology,
Chinese Academy of Science, for Distributed Computing System, Big Data,
Architecture and Machine Learning.

ANALYSIS OF ENTERPRISE SHARED RESOURCE INVOCATION SCHEME BASED ON HADOOP AND R

More Related Content

What's hot (17)

Similar to ANALYSIS OF ENTERPRISE SHARED RESOURCE INVOCATION SCHEME BASED ON HADOOP AND R (20)

Recently uploaded (20)

ANALYSIS OF ENTERPRISE SHARED RESOURCE INVOCATION SCHEME BASED ON HADOOP AND R