close

Вход

Забыли?

вход по аккаунту

?

978-3-319-67071-3 19

код для вставкиСкачать
Collaborative Filtering-Based Matching
and Recommendation of Suppliers
in Prefabricated Component Supply Chain
Juan Du1,2 ✉ and Hengqing Jing1,2
(
)
1
2
SHU-UTS Business School, Shanghai University, Shanghai 201800, China
dujuan_1031@hotmail.com
Shanghai University and Shanghai Urban Construction (Group) Corporation Research
Center for Building Industrialization, Shanghai University, Shanghai 200072, China
Abstract. In the past 20 years, with the continuous growth of the prefabricated
component supply chain, the integration of fragmented information in the supply
chain has aroused wide attention. At present, the information of all aspects in the
supply chain is isolated, and the problem of the separation of each ring is serious,
which not only results in the isolated decision-making of the parties and the waste
of resources, but also lead to inefficient supply chain. B2B come into being, which
provides real-time data and information interaction for the parties in supply chain,
and improve the overall efficiency of the supply chain. This paper focuses on the
problem of supplier matching, in B2B platform, proposing a collaborative
filtering recommendation algorithm based on matching suppliers, which recom‐
mend suppliers for the buyers accurately and improve the overall efficiency of
the prefabricated construction industry supply chain.
Keywords: Prefabricated component supply chain · B2B platform ·
Collaborative filtering · Clustering analysis · Recommendation algorithm
1
Introduction
1.1 Research Background
The industrialization of construction concerns the prefabricated component supply chain
which converges product design, procurement requirements for production and
processing, logistics services after determining the buyer demand. The construction
industrialization contributes to the implementation of environmental friendly building
materials, and realize the goal of sustainable development [1]. This requires that the
information on the corresponding links in the supply chain must realize real-time
communication and timely adjustments to meet the market demand. The use of B2B
economic sharing platform make the supply chain nodes reach a high degree of coop‐
eration in the direction of development.
Considering the long-term interests of enterprises, the research breaks the traditional
supply chain management mode of the enterprise, and then the collaborative filtering
© Springer International Publishing AG 2018
J. Abawajy et al. (eds.), International Conference on Applications and Techniques
in Cyber Security and Intelligence, Advances in Intelligent Systems and Computing 580,
DOI 10.1007/978-3-319-67071-3_19
Collaborative Filtering-Based Matching and Recommendation
129
and clustering is applied to recommend purchaser suppliers whom the neighbor
purchaser preferred in supply chain management in order to improve the efficiency of
the whole supply chain and reduce the total cost of the supply chain. This paper selects
the current relatively mature technology of collaborative filtering and clustering based
on user behavior analysis method to optimize a recommendation engine for B2B
economic sharing platform green Newell. The platform can recommend the suppliers
who have a cooperative relationship with buyers who have similar preferences for goods
with the current buyer.
This paper puts forward the background and information interaction problems of
prefabricated construction industry supply chain. Based on the B2B platform, the
problem of supplier matching is put forward. Focusing on the collaborative filtering
recommendation algorithm based on supplier, experimental results showed that when
the supplier personalized recommendation algorithm realize after the completion of
training, the result is fast and the efficiency is obviously improved.
1.2 Literature Review
Bin [2] defines the prefabricated construction as an architectural form that prefabricated
components are firstly processed by the manufacturer in the factory, and then the
building materials are transported to the site to be assembled according to the actual
situation to achieve the required prefabricated blueprints. Prefabricated construction
industry supply chain is a typical supply chain which plan the project implementation
schedule according to the purchaser order. The supply and demand matching process of
the traditional prefabricated supply chain is a kind of cooperation and operation under
the limited information and limited participation, which is difficult to achieve optimal
allocation and coordination. B2B collaborative platform based on collaborative filtering
provides a way to solve the above problems, provide the buyer the best dynamic match
vendor results and make a reasonable forecast for future orders.
Wang [3] proposed personalized recommendation through collaborative filtering
algorithm, aiming at recommending suppliers to target buyers according to the neighbors
who have similar preferences to purchaser. The core problem for the collaborative filtering
algorithm is the optimal matching of suppliers, in the selection of suppliers, some scholars
have made related research: Pazzani considered that the social attribute similarity infor‐
mation for individual users can reflect the similarity of users interested in the purchase.
He put forward to fill the value on the basis of the data using the social attribute informa‐
tion individual users, but this method may infringe personal privacy information [4].
Cheung proposed the application of Web data mining and data analysis of the server log,
and then proposes a recommendation method of based on the data of implicit evaluation.
This method is found to contain a wealth of information hidden in the data to supply the
explicit data so as to eliminate the sparsity problem [5]. Liu and Shih [6] proposed three
indicators of the time of purchase, purchase frequency, purchase amount to measure
customer lifetime value so that businesses can find more valuable customers, and then
proposes a recommendation method based on customer lifetime value in order to make the
recommendation more targeted. Kavitha puts forward a pre clustering method for the
users with similar scores. Based on the similarity of the users in the cluster, predicting the
130
J. Du and H. Jing
non-scoring data of the users, and the prediction method has achieved good results [7].
Goldberg [8] was the first one to use collaborative filtering recommendation method in
news and film recommendations, received a high praise in 90s.
Reviewing the existing literature, the research on supplier selection is mainly in the
stage of qualitative analysis, and some scholars put forward some qualitative objectives
to measure customer value. But at present, there is a lack of quantitative analysis of user
evaluation and supplier filtering recommendation methods are mostly used in news and
film industry. In this paper, this paper uses quantitative methods to evaluate the suppliers,
exploring the problem of supplier matching in the prefabricated component B2B plat‐
form, and then a recommendation algorithm based on collaborative filtering is proposed.
2
Supplier Selection in B2B Collaborative Platform
of Prefabricated Components
2.1 B2B Collaboration Platform
Guo [9] defines B2B as the business model that the enterprise make trading activities
through real-time information and data exchange. The basic idea of the B2B platform
faced prefabricated construction industry supply chain is establishing a biggest
economic sharing platform converge consulting, sale, processing, logistics. The plat‐
form aims at increasing the transparency of the market, oriented at the coordination
between the various links in the supply chain, which improve the degree of collaboration
of enterprise in the prefabricated construction industry supply chain and the operation
of the whole supply chain level.
The service provided by the B2B platform faced prefabricated construction industry
supply chain include consulting, sale, logistics, and a series of electronic business func‐
tions. Relatively perfect function design compound prefabricated construction industry
supply chain, and improve the efficiency of supply chain management. Secondly,
management integration of the prefabricated construction industry supply chain over‐
come geographical barriers, so buyers will not give up the purchaser whose location is
not convenient. Finally, the integration of the supply chain also reflects the company’s
strategic level, tactical level and the operational level of collaborative services, and
realize information seamless sharing on the B2B platform.
2.2 Supplier Selection Problem
Considering the particularity of the prefabricated construction industry, not all suppliers
can meet the request of the buyers, and the service provided by different suppliers is
different. So the potential customers of every supplier is not the same. In the same way,
buyers who prefer to take a more proactive stance are also different in their preferences
for suppliers. In this paper, based on the above information for both suppliers and buyers’
demand and psychological status, we put forward the following questions and try to use
the method of collaborative filtering to solve it: how to establish a supplier recommen‐
dation system in numerous suppliers to help suppliers find their potential users
Collaborative Filtering-Based Matching and Recommendation
131
successfully, and help buyers find suppliers to meet their needs, achieving a “win-win”
situation. In order to realize the collaborative filtering recommendation algorithm, the
following three steps are indispensable: (1) Collecting the score data of buyer for
suppliers to avoid the problem of the supplier’s recommendation result is not accurate.
(2) Finding similar users and items, and then calculate similarity between the purchaser
and the prefabricated component in an appropriate method. (3) Selecting the user based
collaborative filtering ideas for the supplier personalized recommendation.
3
Collaborative Filtering-Based Supplier Personalized
Recommendation
3.1 Experimental Design of Collaborative Filtering-Based Supplier Personalized
Recommendation
The concepts and methods involved in the experiment are as follows
(1) Collaborative filtering algorithm via Purchaser–Supplier rating matrix
Here a denotes the total number of purchasers registered on the B2B e-commerce
platform, and Ai represents each purchaser, of which i = 1, 2,…a. Similarly, b denotes
the total number of suppliers, and Bj represents each purchaser, of which j = 1, 2, …b.
The specific supplier evaluation system is set as follows:
Ck represents attributes of the suppliers in turn, of which k = 1, 2, 3, 4. The purchaser
evaluates the 4 evaluation indexes of suppliers in order by taking different weight on
the indexes. Dij represents the final result. Evaluation uses round figures within 0 and 5.
The larger the round number, the better suppliers performed (Table 1).
Table 1. Rating matrix AB
A1
A2
…
Ai
B1
B2
…
Bj
D11
D21
…
Di1
D12
D22
…
Di2
…
…
…
…
D1j
D2j
…
Dij
(2) Similarity computation between purchasers - cosine similarity
After data reduction, the next step is to choose a similarity computation method that
gear to the characteristics of prefabricated construction industry, and then generate the
similarity of supplier selection. The production of prefabricated components will vary
from the needs of purchasers to a large extent, and the criteria for each purchaser scoring
for suppliers may be miles apart. Therefore, the unification of measurement standard
should be taken into consideration when computing similarity. In conclusion, cosine
similarity is used to compute similarity in the paper.
132
J. Du and H. Jing
If Xt and Yt are considered as two vectors, the mathematical expression of cosine
similarity is:
∑n
(Xt ∗ Yt)
1
sim(Xt, Yt) = cos  = √t=1
∗√
∑n
∑
n
(Xt)2
(Yt)2
t=1
t=1
(1)
Formula1: Cosine similarity formula
Here Ut denotes vector set according to the comprehensive rating of purchaser for
product, of which t = 1, 2, …i. If U1 = (D11, D12, …D1j),represents the vector composed
by purchaser A1, and U2 = (D21, D22, …D2j), represents the vector composed by
purchaser A2. Then vector U1 and U2 are put into cosine similarity formula to calculate
the cosine of the angle. If the result is closer to 1, then higher similarity between
purchaser A1 and A2 and the closer the preference to the supplier.
(3) Purchaser clustering method – K-means method
Classicality is not the only reason that K-means clustering algorithm is classical in
the study analyzing whether purchasers have similar preference. The most significant
advantage of this algorithm is relatively scalable and efficient when processing large
database. Highly intensive cluster will be generated by collecting and analyzing data
from purchaser demand market as well as supplier market. Different classes and achiev‐
able effect contribute to widely vary in purchasers’ demand for suppliers and their prod‐
ucts, which display the advantages of K-means clustering algorithm. The loop iteration
based on user behavior effectively avoids chance, which makes the supplier recommen‐
dation result more accurate.
3.2 The Sources of Experimental Data
The platform automatically stores each purchasing record of the purchaser into the
database, where key attributes are the names of purchasers and suppliers, deal price of
products, material properties of synthetic products, product categories, and the names
and sales volumes of the product, etc. These are objective data that can be obtained from
the previous transaction records. The platform then automatically sends a questionnaire
to purchasers after they confirm receipt of ordered products and ask them to score one
by one according to the 7 indexes. Then it transforms the qualitative satisfaction score
into quantitative 1–5 points, which is easier to obtain a comprehensive score of the
supplier. The result is also stored in the database after purchasers complete scoring for
later recommending suppliers that in the same clustering center.
As for getting the purchaser preference vector for suppliers through the platform,
history searching record based on user browser platform is the way to attain these data
after matching the username and the passwords on login interface. The historical data‐
base contains the times each type of product browsed; the number of clicks purchaser
views the favorites; the volumes of each type of product sold. The scoring of the products
can be obtained through purchasers’ feedback after each purchase, which can decide the
purchaser preference for products that tend to buy and then generate arrays of purchaser
Collaborative Filtering-Based Matching and Recommendation
133
preference vector. Platform database collect and process data to realize supplier person‐
alized recommendation based on collaborative filtering and clustering analyze of user
behavior efficiently and accurately, which is a great improvement compared to tradi‐
tional supplier recommendation mechanism. In the restriction of time and energy, we
use emulated data to simulate the purchaser historical behavior when conducting experi‐
ment.
3.3 Experimental Process
In the phase of data collection, the experiment plans to collect the raw simulation data
from green Newell platform, including the names of the purchasers and suppliers, the
browsing history of each purchaser, each purchaser’s recorded history of the collection
and purchase of goods, the comprehensive score of each purchasing experience. The
specific way to obtain the comprehensive score of goods is to invite users of B2B
economy sharing platform to fill in the goods satisfaction questionnaire after every
transaction. The content of recommend questionnaires is based on product-related
attributes, each indicator being allocated a certain amount of weight with the weighted
average method of the comprehensive score of each purchase. Han [10] allocates the
weight of quality by 23%, the safety coefficient by 28%, integrating degree by 18%, the
price by 11%, delivery cycle by 5%, after-sales service by 10%, material by 5%.
After processing the original data, a perfect product information database of prefab‐
ricated parts needs to be established. Because the prefabricated building accessories
have high standard requirements, so classification database of the prefabricated compo‐
nent should be built so as to obtain data and query conveniently. For example, relevant
parts data samples is accessible through the professional committee of the prefabricated
component of Shanghai municipal engineering construction quality management asso‐
ciation. The code of products of different specifications is identified for the convenience
of reference as follows (Table 2):
Table 2. Prefabricated product attribute database sample
Name
Specification
Steel content ( kg/m3) Code
3200 130
01A
Unit Price
PC prefabricated exterior panel Rinse concrete m3
Ceramic tiles m3
3700 130
01B
PC prefabricated exterior panel Rinse concrete m3
Ceramic tiles m3
3850 130
02A
4550 130
02B
PC prefabricated balcony board Rinse concrete m3
Ceramic tiles m3
3400 160
03A
3650 160
03B
3
3400 160
04A
3
3650 160
04B
3
Rinse concrete m
Rinse concrete m3
3700 260
05A
3800 240
06A
3
3750 125
07A
PC Precast hollow slab
Rinse concrete m
Ceramic tiles
PC Precast girder
PC Precast beam
PC Precast column
m
Rinse concrete m
134
J. Du and H. Jing
After prefabricated product attribute database is established, clustering analysis on
user behavior data is carried out according to the history of the purchaser, and the specific
indicators are set to (1) the times of each type of product browsed by each purchaser (2)
the times of each type of product in the favorites browsed by each purchaser (3) the
times of each type of product bought by each purchaser (4) the comprehensive score for
this product from the purchaser after each transaction. The product preference score on
each purchaser is calculated on the data above by giving weight. This experiment adopts
the pairwise comparison method to give the weight, the four indicators are used to set
the product preference of the purchaser according to the order marked as A, B, C, D,
given the scale and importance, and the contrast results are shown in Tables 3, 4 and 5,
based on the results from seven prefabricated construction related experts in the field of
investigation statistics. These experts come from all parties in prefabricated component
supply chain, involving designer, manufacturer, assembler and so on. After setting up
the weight of each index, because the unit type used to measure each index is different,
standardize the four index to avoid inaccurate recommendation results. After standard‐
ization, the four indicators can be successfully converted to one in order to gain the
comprehensive scores of a product in the database to purchaser, categorizing the buyers
to form a set of buyers history behaviors feature vector T. If there were i purchaser, label
N items in the database according to the order in A ~ O, Ti = (Ei1, Ei2, …, EiN), such as
characteristic of vector purchasers for T1 = (E11, E12, …, E1N), characteristic vector of
No.2 buyer for T2 = (E21, E22, …, E2N), and so on.
Table 3. Pairwise comparison method questionnaire
A
B
C
D
A
1
3
7
9
B
1/3
1
5
7
C
1/7
1/5
1
3
D
1/9
1/7
1/3
1
Table 4. Pairwise comparison method - column standardization
A
B
C
D
A
1/20
3/20
7/20
9/20
B
1/40
3/40
15/40
21/40
C
5/152
7/152
35/152
105/152
D
7/100
9/100
21/100
63/100
Table 5. Pairwise comparison method - the weight
Evaluating factor
Ultimate weight
A (%) B (%)
4.44
9.03
C (%)
29.13
D (%)
57.40
After completing feature vector, that is, determining the history preference data of
each purchaser, based on the core concept of the k-means clustering algorithm, randomly
Collaborative Filtering-Based Matching and Recommendation
135
select 10 characteristic vector T from i purchasers as the initial clustering centers. And
then the rest of the every feature vector T and ten classes of the initial clustering center
vector should be compared one by one into the cosine similarity formula, sorting the
result points into the highest category of similarity according to the comparison. After
the complete round, all the rest of the (i − 10) purchasers is corresponding to the 10
randomly-selected initial clustering centers. However, due to the ten original clustering
center being randomly selected at first, lacking representativeness and typicality, the
final experimental results need to be adjusted.
Adjustment method employs simple and effective weighted average way, and
specific operation is to generalize the preliminary results to the same class in the
purchaser’s demand preference on the characteristic vector of N component scores
weighted average. And treating a new feature vector as the new clustering center in the
current category, then feature vectors of all i purchasers and the characteristics vectors
of ten new clustering centers should be compared one by one into the cosine similarity
formula according to the comparison results points into the similarity of the highest
category. Calculating the same clustering center and weighted average of the four indi‐
cators score to get a new set of feature vector clustering center. The rest can be done in
the same manner until termination conditions appear.
After implementation of clustering based on user behavior, goods score matrix of
the purchaser is needed to recommend the suppliers to current purchasers. Concrete
implementation method is to find the current buyers of clustering center, add up the score
of all products from all other buyers to obtain the purchaser’s composite score of all the
goods on the platform. The purchaser in accordance with the requirements select the
component from the database, and the platform select the products meeting the require‐
ments, according to accumulation of high and low scores corresponding supplier ranking
for the current buyers personalized recommended suppliers. It has to be based on the
current buyers used by other buyers in the clustering center is located, but the current
buyers never cooperation supplier of comprehensive score as the final supplier person‐
alized recommendation based on a recommendation to the purchaser. So a collaborative
filtering supplier-personalized recommendation algorithm can produce a suggestion list
supplier in the descending order for each purchaser, which is the result of clustering
analysis based on user behavior, has a certain degree of accuracy.
3.4 The Realization of the Experiment
This paper uses the programming software Python to implement the supplier’s person‐
alized recommendation algorithm based on collaborative filtering and user clustering
designed in this experiment. Assuming that the B2B e-commerce platform registered a
total of 1,000 buyers and a total of 2,000 products in the database, firstly you need to
simulate the program through the 1000 buyers’ historical behavior data. In the experi‐
ment, in order to control the number of times, the number of visits, the number of visits
after collection and the number of purchases are set to the rank number of 1–10. And
the actual number is set to the rank number that multiplied by 100. The rule is the number
of visits> = the number of visits after collection> = the number of purchases, and the
number of times must be greater or equal than 0. Then, in the process, these four
136
J. Du and H. Jing
indicators are given the weight respectively that is calculated by the two pairs of analysis
in order to complete the 1000 buyers’ feature vectors.
With each buyer’s feature vector, the user clustering analysis can be performed. The
condition of the loop termination is set that after another K-means clustering algorithm
has no object to be reassigned to different clusters. The results are shown in the table of
user Clus (the code of cluster analysis is shown in Fig. 1).
Fig. 1. The Python code of cluster analysis
After completing the cluster analysis, the next step of the experiment is to recom‐
mend the supplier to the current purchaser based on the idea of collaborative filtering
and the analysis of user’s behavior. The specific programming idea is to accumulate the
comprehensive score of each product for all buyers who belong to the same clustering
center. The score of 0 indicates that the buyer has not made a deal with the supplier.
And the final score of the results stored in the table of productClus to be used to train
the supplier personalized recommendation algorithm (collaborative filtering recom‐
mendation algorithm is shown in Fig. 2).
The recommended method is to find other buyers who are at the same clustering
center as the current purchaser. After excluding the suppliers who have had a partnership
with the current purchaser, the method recommends the suppliers who have the same
preference to the current buyer according to the level of cumulative level. When designing
the process, the paper does not consider that the prefabricated construction industry
buyers often have clear demand. But if there is a database table, The SELECT statement
in the SQL statement can be used to filter out the prefabricated components that the
current buyers needed. This procedure only implements a vendor recommendation algo‐
rithm based on collaborative filtering. The program’s search variable is set to the current
suppliers’ number and the number of recommended suppliers. If recommending the top
Collaborative Filtering-Based Matching and Recommendation
137
Fig. 2. The Python code of collaborative filtering recommendation
10 suppliers for the 68th buyer, (68, 10) need to be input in the program. The results of the
operation of the program are: [1435,374,427,1057,795,1365,1936,769, 1377,1727]. Then
the suppliers who corresponds the product that are ran out from the results of the program
should be recommended to the current buyers.
3.5 Analysis of the Results of the Experiment
Through the experimental results, it is found that when the supplier’s personalized
recommendation algorithm is trained, the recommended results are quickly and accurate
based on collaborative filtering and user’s historical behavior. Compared with the tradi‐
tional model, only using the content recommendation to recommend the items that is
similar to the users’ previous favorite items for them, the efficiency is significantly
improved. The supplier’s personalized recommendation algorithm is based on the same
part and the different entirety to match the current buyers to the supplier. While saving
manpower and resources at the same time, the prefabricated construction industry supply
chain management will be significantly improved as a whole, which lays the foundation
to introduce the platform to the industry.
Because of the collaborative filtering and user behavior analysis technology is rela‐
tively mature, the accompanying problems are fixed and obvious in this experiment: (1)
the cold starting problem: collaborative filtering technology is mainly based on the user’s
historical score of the project. When the score sources are insufficient, that is difficult
to make accurate recommendations. As for e-commerce systems, there is a large number
of new users accessing and adding the new projects every day. The system only works
138
J. Du and H. Jing
effectively for new users and new projects to better retain the system for customers and
dig the potential customers. (2) The data sparse problem: in practical application, the
user generally only can evaluate (or buy) a small number of items. The scoring matrix
is generally very sparse. In this case, the challenge is to get accurate predictions with
relatively few effective scoring. The main idea is to use the assumption of the user’s
taste and then increase the additional information matrix.
4
Summary
In this paper, a relatively mature concept of collaborative filtering recommendation
algorithm is applied to a new field, which is the prefabricated construction industry’s
B2B economic sharing platform. And as much as possible, each step would choose the
methods that match the feature of the industry through the comparison of the methods
and methods into the experiment. The results of the experiment show that the design of
the supplier personalized recommendation algorithm have a significant effect, but
whether in B2B platform’s design or the implementation of the design still have some
problems that need the further study in a relatively new field to achieve a stable effect.
Based on the research results of this paper, the author suggests that we can also
proceed with this analysis from the following aspects: (1) in this paper, we discuss the
new supply chain collaborative management of prefabricated construction industry’s
B2B economic sharing platform’s content, strategy and platform basic functions, hoping
to establish the corresponding prefabricated construction industry supply chain collab‐
orative performance evaluation model in the following study. People can use the model
to assess the supply chain nodes of each company’s ability of synergies. (2) This paper
only focuses on how to design experiments for the current buyers to recommend the
right suppliers. But this paper did not consider the economic environment and the needs
of buyers, which is not static. So if you want to occupy the market for a long time, the
platform developers must achieve the function of the market forecasting. That is using
the data in previous years for the needs of buyers or the number of platform orders to
make a reasonable forecast to avoid a serious imbalance relationship between supply
and demand situations. And then, showing the advantages of e-commerce platform is
also worthy of the further study.
Acknowledgement. This work was supported by Natural Science Foundation of Shanghai
Project under Grant 15ZR1415000.
References
1. Zeng, T.: Research on digital lean construction platform based on supply chain. Civ. Eng.
Inf. Technol. 5(5), 34–39 (2013)
2. Li, B.: The present situation and development of prefabricated buildings in China. Eng. Sci.
Technol. China Inf. Technol. 7, 114–115 (2014)
3. Wang, H., Nie: Scalable collaborative filtering algorithm based on fuzzy clustering. Online
Sci. Pap.
Collaborative Filtering-Based Matching and Recommendation
139
4. Pazzani, M.J.: A framework for collaborative, content-based and demographic filtering. Artif.
Intell. Rev. 13(5–6), 393–408 (1993)
5. Cheung, K.W., Kwok, J.T., Law, M.H.: Mining customer product ratings for personalized
marketing. Decis. Supp. Syst. 35, 231–243 (2003)
6. Liu, D.R., Shih, Y.Y.: Hybrid approaches to product recommendation based on customer
lifetime value and purchase preferences. J. Syst. Softw. 77(2), 181–191 (2005)
7. Kavitha, M.K.: Kernel based collaborative recommender system for E-purchasing. Acad. Sci.
35, 513–524 (2010)
8. Goldberg, D.: Using collaborative filtering to weave an information tapestry. Commun. ACM
35(12), 61–70 (1992)
9. Guo, J., Jinzhao, Wang, Z.: Research on the evolution of online supply chain financial model
and risk management. Bus. Econ. Manag. (2014)
10. Han, J.: Research on intelligent selection and evaluation system of supplier based on fuzzy
comprehensive evaluation, p. 35. Dissertation of Shandong University (2014)
Документ
Категория
Без категории
Просмотров
1
Размер файла
931 Кб
Теги
978, 67071, 319
1/--страниц
Пожаловаться на содержимое документа