Nâng cao hiệu quả dự báo nhu cầu cho các doanh nghiệp dựa trên ứng dụng công nghệ kinh doanh thông minh (BI)
Kỷ yếu Hội thảo quốc tế “Thương mại và phân phối” lần 2 năm 2020
ENHANCING THE EFFECTIVENESS OF DEMAND FORCAST FOR BUSINESSES
BASED ON BUSINESS INTELLIGENT
NÂNG CAO HIỆU QUẢ DỰ BÁO NHU CẦU CHO CÁC DOANH NGHIỆP DỰA TRÊN ỨNG
DỤNG CÔNG NGHỆ KINH DOANH THÔNG MINH (BI)
Hoang Ngoc Canh, MA, Trinh Thi Nhuan,MA, Nguyen Thi Thu Thuy,PhD
Thuongmai University
Abstract
Demand forcast plays an important role in determining the failure or success of enterprises as effective
demand forcast may help with minimizing the cost, reducing inventory, providing the most suitable products for
customers, and enhancing competitiveness of enterprises, etc. Currently in the world, different tools have been
applied to forcast businesses’ demands, in which, a trend of using business intelligent applications based on
information technology (IT) and artificial intelligent (AI) has been in favour. In Vietnam, the idea of business
intelligent (BI) is still not clear. There is only a small number of businesses have been applying BI into their
performance in general and in forcasting demand in particular. The paper focuses on some limitations in
forecasting demand of Vietnamese enterprises. Besides, the author would like to introduce some basic information
of BI and analyse the application results of a typical BI tool in forcasting demand of a certain enterprise.
Keywords: business intelligent, demand forcast, forcast…
Tóm tắt
D ự báo nhu cầu đóng vai trò quan trọng quyết định đến sự thành công hay thất bại của các doanh
nghiệp. Bởi dự báo nhu cầu chính xác sẽ góp phần giảm thiểu chi phí, giảm hàng hóa tồn kho, cung cấp sản
phẩm phù hợp nhất cho khách hàng, tăng sức cạnh tranh của doanh nghiệp… Trong cuộc cách mạng công
nghiệp 4.0, việc áp dụng công nghệ thông minh trong sản xuất, kinh doanh là điều không thể thiếu. Hiện nay,
trên thế giới có nhiều công cụ khác nhau được áp dụng trong dự báo nhu cầu của doanh nghiệp, trong đó xu thế
nổi bật là sử dụng các công cụ ứng dụng công nghệ kinh doanh thông minh dựa trên nền tảng công nghệ thông
tin và trí tuệ nhân tạo. Tại Việt Nam, khái niệm kinh doanh thông minh (Bussiness Intelligent - BI) còn khá mờ
nhạt, chỉ có một số lượng nhỏ doanh nghiệp đã và đang ứng dụng công nghệ này trong hoạt động kinh doanh
nói chung và trong dự báo nhu cầu nói riêng. Bài viết tập trung vào một số hạn chế trong dự báo nhu cầu của
các doanh nghiệp Việt Nam, giới thiệu đôi nét cơ bản về BI và đưa ra những phân tích về kết quả ứng dụng một
công cụ BI điển hình trong dự báo nhu cầu trong doanh nghiệp cụ thể. Từ đó độc giả sẽ hiểu được về sự hữu ích
khi sử dụng BI trong dự báo, nhằm nhân rộng việc ứng dụng BI trong quản trị kinh doanh.
Từ khóa: công nghệ kinh doanh thông minh; dự báo; dự báo nhu cầu
1. Vietnamese businesses and limitations on forecasting demand at present
Demand forecast is one of the biggest challenges that businesses often encounter when operating
in a developing market that has been greatly affected by internal factors, customs, culture as well as the
unstability of the macroscopic policies in Vietnam. In fact, most Vietnamese businesses have been
making demand forecasts by using forecast features of integrated business management software systems
(e.g.: ERP, CRM) or even Office utilities like Excel. [10] These utilities are not specifically designed for
forecasting as they are mainly used for managing, storing, and processing. The use of in-depth
algorithms to explore and analyse the data in predictive manners is not emphasized, resulting in
inaccurate forecasts as well as hard to apply to high level real-world requirements. Besides, many
Vietnamese businesses have faced difficulties in thoroughly understanding the process of forecasting,
collecting, standardizing and processing data, choosing appropriate forecast approaches based on the
available resources. They have even never heard of or accessed any forecasting tools or functions in
business operation. The consequence of these issues is that a significant proportion of firms have not yet
made effective forecast, or done it by emotion and experience rather than using statistical methods based
on data. Althought the other have had conerns about demand forecast, they have only taken advantages
831
Kỷ yếu Hội thảo quốc tế “Thương mại và phân phối” lần 2 năm 2020
of integrated forecast functions available in management tools and forgot about a good investment in
human resources and specialized systems for forcasting.
One cause to the shortcoming of the forecasts is the levels of details. There are businesses
choose to make overall forecasts by regions, categories, ranges, or brands of products leading to the
lack of thoroughness. That is the reason why the forecast activities cannot effectively support plans for
purchasing or production. Vice versa, there are also firms that focus on forecasting at most detailed
levels such as commodity codes because of lacking tools and methods that can detect and synthesize
the relevant information between commodity codes, turning into conflicts in making decisions among
forecast results. Businesses therefore find it difficult to make certain decisions on each item.
Nowadays, specialized tools to forecasting like FuturMaster, ForecastPro have been used by
countries in the regions or worldwide and shown their effectiveness of 85% to 95% accurate cases.
These applications share a common platform called Business Intelligent (BI). Being a high-end and
professional solution, BI can not only applied for large corporations but also be a decisive support
solution for businesses of sizes and different operation fields. In application, BI obviously best
facilitates businesses in the fields of consumer goods, beverage, and foods where time factor is of
priority. Globally, BI has become a frequent management tool of many high-profiles like Coca-Cola,
Unilever. In Vietnam, only a small number of big names have been deploying BI and considered it as a
secret weapon while most others still do not really approach or worry about its efficiency over the cost
of deploying these systems.
This situation may be attributed to these following reasons: (1) the foundation and environment
of application development technology in Vietnamese enterprises are always slower than the world;
(2) enterprises have not had a roadmap or been aware of the preparation for the application of
technology platforms for the present as well as the future. It can be seen clearly through the
rudimentary process of storing, linking sources of data in businesses which mostly includes documents
or simple digital files such as Excel, small and seperated database, the underestimation of old data,
says, the most important sources of input materials for the advanced systems, technologies and
algorithms applied for demand forecast.
2. Solutions to enhancing demand forecast for businesses based on business intelligence
2.1. An overview of Bussiness Intelligent (BI)
There are a number of definitions of BI, each identifies a prominent feature of BI. However,
they all refer to BI’s ability to help drive business decision-making process. BI is therefore known as a
solution to smart business management, a report system enabling organizations, businesses to access
data from different sources about customers, markets, providers, clients, personnel, etc.; analyse and
use such data as meaningful sources of information to support predictions and decisions. [1]
In fact, organizations and businesses need BI for data integration and information analysis. For
managers, BI is the most accurate and comprehensive analytic system as its information is aggregated
from multiple sources in business performance. In many cases, if BI is out of choice, it may take
organizations and businesses long time to get results and even cost them a sum for conducting surveys
and researches. Using BI helps businesses have immediate analytical management information to
answer such questions like: “Who is the most important clients of the business?”, “Which market is
bringing the main proportion of profit?” [8]
In short, BI can be considered as the optimal solution to the problems of forcasting, decision-
making support, queries and reports, online processing and analysis, statistical analysis, and data
mining. Objects interacted with BI include management, decision makers, customers, and analysts.
2.2. The benefits of business intelligence technology for demand forecast.
How can businesses move in the right direction and develop best? BI, with many tools to help
completely change the business management system, is the answer to the question.
832
Kỷ yếu Hội thảo quốc tế “Thương mại và phân phối” lần 2 năm 2020
In real world, it is complicated to make forecasts for production in a factory; sales for a
commercial enterprise; the compilation of next quarter or next year plans; or the direction of the
company for many years to come. Almost all plans and forecasts for production and sales of
enterprises are based on subjective and qualitative assessments of some experienced ones, especially
businesses applying forecast methods like expert methods, management board’s consultative methods,
customer survey methods, etc. Executives would all probably want reliable and scientific supports that
facilitate making effective forecasts. Understanding this need, the market leaders of BI including
Business Objects, Cognos, and SAP Business Intelligence all provide products that help forecast and
plan effectively for the business. In addition to standard algorithms and procedures, BI is of skillfully
combination with users’ experience (by allowing additional users’ parameters during forecasting) and
can aggregate the results of demand forecasts, or plans quickly and with high level of accuracy.
Besides, BI system can help users analyze hypotheses. This feature allows users to simulate a number
of events, thereby evaluating the changing trend of the KPIs in which they are interested. For the
scientific and accurate forecasting results, BI tools have utilized many of their advantages to analyze
the data as follows: [4]
‐ Centralized data mining: When businesses operate effectively, expansion across many
provinces or countries is indispensable. That leads to the difficulty of managing the dispersed data,
which is usually in various formats, of the company or organization. This process is time consuming
and needs labor intensive whenever a query, analysis and comparison of information needed. With
BI’s Data Warehouse, important data that are scattered across the company and in different formats
can now be regularly extracted and aggregated into unified structure. This process assures the accuracy
and timeliness of detailed to overview reports across the organization. The Data Warehouse has been
widely recognized by many large corporations as an important part of their globalization.
‐ Advanced analysis report: One of the greatest fear of corporate governance is chaotic data.
To organize and manage this source of data is always complicated, let alone exploiting its values.
However, the process of making forecast still requires complex needs. At present, BI's advanced
analytical reporting solution is relatively complete with outstanding features such as:
¾ Data mining to the maximum: BI helps to deal with even complicated requirements such as
“providing information on sales and the number of items sold in the last 3 years, by all regions, applied
to all ranges of products, each product, and staff performing transactions.” With all these types of
questions, it takes the administrator only a couple of seconds interacting with the OLAP system to
have the answers.
¾ Dimensional information customization: Beside the feature of data mining comes the ability
to customize the order of information dimensions. For example, with the same information dimensions
above, we have another perspective such as “providing information about sales and the number of
items sold, for sales staff, of all regions, on all product groups and each product, in the last 3 years.”
‐ Automatic monitoring and alerting: In order to further highlight the names and images in BI
markets, leading providers like Microsoft, BusinessObjects, Cognos, Hyperion, SAS continuously
invest in user interface. The definitions of Dashboards, Scorecards, etc. have been applied to business
management. Thanks to the dashboards, the Key Performance Indicators – KPIs – can be regularly
generated and updated. In addition to automatic coloring and image alerts, the BI system also
automatically sends e-mails to the appropriate authorities, giving managers the ability to get
information about what is going on.
From all of the above, BI has shown its prominent advantages to forecasting in businesses. In
short, it can be briefly understood as follows:
¾ Compared to a traditional tools like Excel: with a set of statistic data on customers, Excel can only
analyze what kind of products that customers used to buy, quantity, and the entrepreneur himself
has to rely on his own experience, knowledge, and skills to analyse the data to make forecasts for
833
Kỷ yếu Hội thảo quốc tế “Thương mại và phân phối” lần 2 năm 2020
a product or service; while BI tools use artificial intelligent (AI) to forecast trends of product
consumption in the future, or what products the prospective customers are likely to be in favour
of.
¾ Assess important features (factors): e.g., assess the factors indicators like income, habit, living
environment, cuture, preferences, shapes, marital status, etc. to find out which is important to the
buying decisions.
¾ Detect groups of mutually dependent attributes.
¾ Detect foreign sets of values, for example, finding fault sets in an unlimited number of records.
2.3. The structure of BI
Figure 1. The elements of BI [5]
2.4. The application of BI to demand forecasting in businesses based on specific data samples
Basic steps to build forecast pattern applying BI.
- Step 1: Prepare sources of data. These can be a collection of statistics records of enterprises
ever and now used for training and testing of the model.
- Step 2: Group the data: the data will be divided into 2 groups: Group 1 is training data and
group 2 is for testing data.
- Step 3: Select models for data mining or data analysis (using available models) provided by
large technology companies based on credible algorithms, e.g. Decision Tree, Classity, Cluster,
Associate, etc. and train using this models on sets of traning data.
- Step 4: Test the above models using testing data sets. Evaluate the test results through tools
such as Lift Chart or Confusion Matrix. If the test results are not satisfactory, change the training data
sets or change the data mining-data analysis model.
- Step 5: Apply the tested models to the real data to make demand forecast for businesses.
Next, the author would like to present a specific application on a hypothetical problem as follows:
Hypothetical problem:
This is a typical case in the world of applying algorithms on data mining-data analysis, namely
using the Data Mining and Analyze Add-ins for Microsoft Excel to forecast prospective customers for
a given Bicycle business company - Adventureworks.
834
Kỷ yếu Hội thảo quốc tế “Thương mại và phân phối” lần 2 năm 2020
This company has a historical data set of customers who have purchased their products as
follows:
‐ 10,000 customers who have purchased with the company (the status of buying or no purchase
has been clearly stated.)
‐ Each customer has 13 attributes: customer ID, marital status, gender, annual income, having
children or not, educational background, profession, home owning, car owning, work travel distance,
residence, age, bike buying decision.
‐ In which, the last attribute is Bikebuyer = Yes/ No. This attribute describe whether the
customer would buy a bike or not.
‐ The aim of the hypothesis is that if there was a new customer data set, (exactly having the
information of 77 completely new customers, not duplicates of the 10,000 customers mentioned above)
and certainly we do not know whether they would buy or not. The thing is how to classify and predict
who is going to buy (Bikebuyer = Yes) and who is not (Bikebuyer = No). In order to solve this task, we
have to build a reliable enough forecast model based on the data source of 10,000 customers recorded.
Solving the problem by applying BI through Data Mining and Analyze Add- ins for Microsoft
Excel:
‐ Data Mining and Analyze Add- ins for Microsoft Excel is an Add-in utility integrated in MS
Excel. Microsoft lines of softwares are commonly used for business management in Vietnam, so this
utility helps making use of the facilities. As being easily accessed and deployed, Data Mining and
Analyze Add- ins needs no big investment in technology and money from the investors. Excel is
basically an environment for this add-in to load data from different sources (most businesses store their
information in Database, XML or Excel files, etc.) and performs mining analysis. This add-in has
versions corresponding to MS Excel 2007 and 2010, in which the author will use the utility for the
2010
version,
specifically
free
download
link
us/download/details.aspx?id=29061.
‐ Data source for the tool: The data set for 10,000 transactions and 77 new customers needs to
be stored in the Data Sheets of any MS Excel file.
‐ Apply Data Mining and Analyze Add-ins for Microsoft Excel to the MS Excel file mentioned
above. The interface of this Add-in is integrated as a function tab as shown below:
Figure 2: Data Mining and Analyze Add- ins for Microsoft Excel
Source: Results of running data from the software
‐ From the original customer data (10,000 records), create two training data sheets (70%
equivalent to 7000 records) and Testing Data (30% equivalent to 3000 records).
‐ Using the training data generated to build the forecast model. This example uses the decision
tree model (a well-known algorithm for data mining techniques ).
‐ Model Validation: After using the Training Data for modeling, we now use Testing Data to
check if the model is good enough to use. (If it is not good, use another Training Data). There
are three main techniques for testing the model: Accuracy Chart (Lift Chart), Classification
Matrix and Profit Chart.
835
is t
Kỷ yếu Hội thảo quốc tế “Thương mại và phân phối” lần 2 năm 2020
‐ Here are some results after using Lift Chart and Classification Matrix.
Figure 3: Test Results via Lift Chart
Source: Results of running data from the software
‐ Figure 4 shows a chart with the X-axis is the total percentage of the Data Testing, the Y-axis
he total anticipated percentage of bicycle purchasers corresponding to X (percentage of Y-column is
against the total number of actual buyers in the Data Testing above.) There are three line graphs, in
which the Random Guess is a linear with a 50/50 chance of correctly guessing the buyers. This line
shows the lowest state of the forecasting. In contrast, the Ideal Model is the line repesenting the ideal
status when with at least 10% of Testing Data, the system can accurately predict 100% of bike buyers.
The middle line is (describes the results of the Decision tree model) therefore the target of considering.
It reflects the upward trend of the Y-axis value, which is acceptable comparing to the variation of X.
Geometrically, the further this line is away from Random Guess, the better result it will be. It is
obvious that with a small number of records in Testing Data, the model also tried to calculate to find
the largest number of buyers as possible. For example, there are 3000 records in Testing Data, of
which the number of bike buyers (Bikebuyer = Yes) is 49; Thus, for each of 50% of the data in Testing
Data, the model can find approximately 73% of the buyers (49 * 73% ~ 36 people). This is an
acceptable result for the effectiveness of the model.
Figure 4: Classification Matrix test results
Source: Results of running data from the software
836
Kỷ yếu Hội thảo quốc tế “Thương mại và phân phối” lần 2 năm 2020
From Figure 5, we find out that the total correct forecasts (including correct guesses of buyers
and forecasts of the non-purchasers) accounted for 89.13% (equivalent to 2,764 customers). This is a
clear indication that the model builds a positive response to the test on the Test Data and can be
confidently used in forecasting.
Model Usage:
¾ After being tested (Testing) with acceptable accurate results, the model then can be used to predict
unknown data. We now use the model of Decision Tree built to predict the number of bike buyers.
The predictable data in “New Customers” sheet include 77 customers (note that this customer data
does not have the BikeBuyer attribute because this is a predicted future data and these customers
do not appear in the Training and Testing Data). After processing data, a sheet of listing 77
customers will be generated (the customers’ attributes can be shown optionally – here displays
only ID attribute for example). The Predictive Result column showing Yes/No attributes is
equivalent to the BikeBuyer attribute - this is the forecast of the buyer's ability to buy or not to buy
a bicycle.
¾ Beside predicting the chance of buying or not of customers, the model also specifies if two
customers are expected to buy a bike, customers with a higher probability of buying through the
Predictive Probability will be labeled for Yes value. Consult Figure 6 below:
Figure 5: Predictable results using the model
Source: Results of running data from the software
From the above analysis, applying BI to customer demand forecast, it will be much easier for
businesses to predict the buying/not buying demand of new customers and forecast the percentage of
purchases. This helps businesses have suitable plans and trategies in customer care and right allocation
of human resources to achieve the best results.
3. Conclusion
The 4.0 Industrial Revolution is an indispensable trend to meet the exact needs of each
customer, increasing productivity, quality, efficiency, speed, reducing raw material consumption, and
production and operation costs. This trend includes the protection of the living environment and
gradual development. As an evolution of supply and demand, intelligent business management
solutions have been constantly emerging, among them Business Intelligence (BI) has been developed
and strongly compatible with the abovementioned advanced information management systems. The
paper addressed the questions of demand forecast to businesses, the limitations on demand forecasts,
demonstration of BI’s huge benefits, suggestions of BI application, particularly the Data Mining and
Analyze Add- ins for Microsoft Excel in demand forecast. The author hopes this paper can become an
useful references for organizations and businesses in the era of knowledge competition in the door of
the 4.0 industrial revolution.
837
Kỷ yếu Hội thảo quốc tế “Thương mại và phân phối” lần 2 năm 2020
REFERENCES
1. Nguyễn Thị Thu Thủy, Hoàng Ngọc Cảnh (2017), “Using Business Intelligence for analyzing students’
data in ThuongMai University of VietNam, The Journal of Internet Electrolic Commerce research”, Vol.
17, No 2 (April), pp: 1-14
2. Nguyễn Thị Thu Thủy, Hoàng Ngọc Cảnh (2015), “Using Business Intelligent Tools in Commercial
University of Vietnam”, Kỷ yếu Hội Thảo Quốc gia - MIS của ĐH KT TP HCM
3. Hoàng Ngọc Cảnh (2016), “Công nghệ Business intelligence và Hệ thống thông tin quản lý của tổ chức -
doanh nghiệp”, Kỷ yếu Hội Thảo Quốc gia - Vai trò của hệ thống thông tin quản lý đối với sự phát triển
của tổ chức - doanh nghiệp
4. Chugh, R & Grandhi, S 2013, ‘Why Business Intelligence? Significance of Business Intelligence tools and
integrating BI governance with corporate governance’, International Journal of E-Entrepreneurship and
Innovation, vol. 4, no.2, pp. 1-14.
5. Dedić N. & Stanier C. (2016). "Measuring the Success of Changes to Existing Business Intelligence
Solutions to Improve Business Intelligence Reporting". Lecture Notes in Business Information
Processing. 268. Springer International Publishing. pp. 225–236.
6. Pendse, Nigel (7 March 2008). "Consolidations in the BI industry". The OLAP Report
7. Julian, Taylor (10 January 2010). "Business intelligence implementation according to customer's needs".
APRO Software. Retrieved 16 May 2016
8. Rud, Olivia (2009), “Business Intelligence Success Factors: Tools for Aligning Your Business in the
Global Economy” Hoboken, N.J: Wiley & Sons
9. Data Warehouse và Business Intelligence (Data Warehouse and Business Intelligence ) – Website:
10. So sánh OLTP và OLAP (Compare OLTP and OLAP) – Website: https://viblo.asia/p/oltp-va-olap-co-gi-
khac-nhau-maGK786BZj2
11. Dự báo tốt giúp doanh nghiệp nâng cao hiệu quả kinh doanh
–
Website:
http://baocongthuong.com.vn/du-bao-tot-giup-doanh-nghiep-nang-cao-hieu-qua-kinh-doanh.html/-+
838
8 trang
25240
Free
Bạn đang xem tài liệu "Nâng cao hiệu quả dự báo nhu cầu cho các doanh nghiệp dựa trên ứng dụng công nghệ kinh doanh thông minh (BI)", để tải tài liệu gốc về máy hãy click vào nút Download ở trên
File đính kèm:
- nang_cao_hieu_qua_du_bao_nhu_cau_cho_cac_doanh_nghiep_dua_tr.pdf
