7

KHO DỮ LIỆU VÀ HỆ THỐNG HỖ
TRỢ RA QUYẾT ĐỊNH

CHƯƠNG 7
MỘT SỐ CÔNG NGHỆ VÀ KỸ THUẬT CHO HỆ HỖ
TRỢ RA QUYẾT ĐỊNH
Bộ môn Khoa học máy tính – Khoa Hệ thống thông tin quản lý
Mục tiêu
1. Hiểu về báo cáo kinh doanh

2. Nhận biết nhu cầu và sức mạnh của báo cáo kinh doanh
3. Hiểu tầm quan trọng của trực quan hóa dữ liệu/thông tin
4. Tìm hiểu các loại kỹ thuật trực quan khác nhau
5. Đánh giá cao giá trị mà phân tích hình ảnh mang lại cho BI/BA
6. Biết các khả năng và hạn chế của bảng điều khiển
Bộ môn Khoa học máy tính – Khoa Hệ thống thông tin quản lý 2
Nội dung
1. Báo cáo kinh doanh

2. Một số kỹ thuật cho mô hình dự báo
3. Một số kỹ thuật ra quyết định dựa trên mô hình
4. Hệ ra quyết định tự động và Hệ chuyên gia
Báo cáo kinh doanh
Tổng quan về báo cáo kinh doanh (Business report)
Báo cáo = Thông tin  Quyết định

Báo cáo?
Một số phần tử giao tiếp được tổ chức để truyền đạt thông tin đặc biệt
Một báo cáo có thể đáp ứng nhiều chức năng
To ensure proper departmental functioning
To provide information
To provide the results of an analysis
To persuade others to act
To create an organizational memory…
Báo cáo kinh doanh là gì?
Một tài liệu được tạo ra nhằm chứa thông tin về tình hình kinh
doanh
Mục đích: để cải thiện các quyết định quản trị
Nguồn: dữ liệu từ bên trong và bên ngoài tổ chức (thông qua
việc dùng ETL)
Hình thức: text + tables + graphs/charts
Phân phối: in-print, email, portal/intranet
 Thu nhận dữ liệu (Data acquisition)  Tạo thông tin
(Information generation)  Ra quyết định (Decision making) 
Quản lý qui trình (Process management)
Business Reporting
Chìa khóa của báo cáo thành công
Rõ ràng (Clarity) …

Ngắn gọn (Brevity) …
Đầy đủ (Completeness) …
Chính xác (Correctness) …
Các kiểu báo cáo (theo nội dung và hình thức)
Phi hình thức (Informal) – a single letter or a memo
Hình thức (Formal) – 10-100 pages; cover + summary + text
Báo cáo ngắn (Short report) – periodic, informative, investigative
Cam kết chất lượng dịch
Các loại báo cáo kinh doanh
vụ (service-level
agreements)
Báo cáo số liệu quản trị (Metric Management Reports)
Help manage business performance through metrics (SLAs for externals;
KPIs for internals)
Can be used as part of Six Sigma and/or TQM Total Quality
Management
Dashboard-Type Reports
Graphical presentation of several performance indicators in a single page
using dials/gauges
Báo cáo dạng thẻ điểm cân bằng (Balanced Scorecard)
Include financial, customer, business process, and learning & growth
indicators
Các thành phần của Hệ thống báo cáo kinh doanh
Các đặc tính chung

OLTP (online transaction processing)
ERP, POS, SCM, RFID, Sensors, Web, …
Nguồn dữ liệu(dung lượng, sự đa dạng, tốc độ, …)
ETL
Lưu trữ dữ liệu (Data storage)
Business logic: cách mà các sự kiện/giao dịch được chuyển đổi thành
metrics, scorecards, và dashboards
Phương tiện phân phối (Publication medium)
Sự đảm bảo (Assurance)
Trực quan dữ liệu và thông tin (Data and Information
Visualization)
“The use of visual representations to explore, make sense of, and
communicate data.”
Data visualization vs. Information visualization
Information = aggregation, summarization, and contextualization
of data
Related to information graphics, scientific visualization, and
statistical graphics
Often includes charts, graphs, illustrations, …
Tóm tắt lịch sử trực quan dữ liệu
Data visualization can date back to the second century AD

Most developments have occurred in the last two and a half
centuries
Until recently it was not recognized as a discipline
Today’s most popular visual forms date back a few centuries
The First Pie Chart
Created by William Playfair in 1801
William Playfair is
widely credited as
the inventor of the
modern chart,
having created the
first line and pie
charts.
Decimation of Napoleon’s Army During the 1812
Russian Campaign
• Arguably the most popular multi-dimensional chart By Charles Joseph Minard

Tóm tắt lịch sử trực quan dữ liệu
1900s –
more formal attitude toward visualization
focus on color, value scales, and labeling
Publication of the book Semiologie Graphique
2000s –
Emergence of Internet as the medium for information visualization 
raising visual literacy
Incorporate interaction, animation, 3D graphics-rendering, virtual worlds,
real-time data feed
2010s and beyond – ?
Các kiểu Biểu đồ và Đồ thị
Which one to use? Where and when?

 Basic Charts and Graphs  Specialized Charts and Graphs
 Line Chart  Histogram
 Bar Chart  Gantt Chart
 PERT Chart
 Pie Chart
 Geographic Map
 Scatter Plot
 Bullet Graph
 Bubble Chart
 Heat Map / Tree Map
 Highlight Table
A Gapminder Chart Wealth and Health of Nations
See
gapminder.org
for interesting
animated
examples
Góc phần tư ma thuật (Magic Quadrant)
Gartner Magic Quadrant là đỉnh cao nghiên cứu trong một thị
trường cụ thể, mang đến cho ta cái nhìn toàn cảnh về các vị trí
tương đối của các đối thủ cạnh tranh trên thị trường. Bằng cách
áp dụng kỹ thuật đồ họa và bộ tiêu chí đánh giá thống nhất,
Magic Quadrant giúp ta nhanh chóng xác định cách các nhà cung
cấp công nghệ đang thực hiện tầm nhìn đã đề ra của họ như thế
nào và cách họ hoạt động tốt như thế nào theo quan điểm thị
trường của Gartner.
(Nguồn:
https://www.gartner.com/en/research/methodologies/magic-
quadrants-research)
Góc phần tư ma thuật (Magic Quadrant)
Các đơn vị hàng đầu

(Leaders)
Các đơn vị có tầm nhìn
(Visionaries)
Các đơn vị thị trường
ngách (Niche Players)
Các đơn vị thách thức
(Challengers)
Sự nổi lên của trực quan dữ liệu và phân tích trực
quan (Visual Analytics)
Magic Quadrant for Business Intelligence and Analytics Platforms

(Source: Gartner.com)
Many data visualization companies are in the 4th quadrant
There is a move toward visualization
Magic Quadrant for Analytics and Business Intelligence
Platforms
Source: Gartner (February 2020)

Sự nổi lên của trực quan dữ liệu và phân tích trực
quan (Visual Analytics)
Emergence of new companies

Tableau, Spotfire, QlikView, …
Increased focus by the big players
MicroStrategy improved Visual Insight
SAP launched Visual Intelligence
SAS launched Visual Analytics
Microsoft bolstered PowerPivot with Power View
IBM launched Cognos Insight
Oracle acquired Endeca
Phân tích trực quan (Visual Analytics)
A recently coined term

Information visualization + predictive analytics
Information visualization
Descriptive, backward focused
“what happened” “what is happening”
Predictive analytics
Predictive, future focused
“what will happen” “why will it happen”
There is a strong move toward visual analytics
Visual Analytics by SAS Institute
SAS Visual Analytics Architecture

Big data + In memory + Massively parallel processing + ..
Visual Analytics by SAS Institute
At teradatauniversitynetwork.com, you can learn more about SAS

VA, experiment with the tool
Performance Dashboards
Performance dashboards are commonly used in BPM software

suites and BI platforms
Dashboards provide visual displays of important information that
is consolidated and arranged on a single screen so that
information can be digested at a single glance and easily drilled in
and further explored
Dashboard design
The fundamental challenge of dashboard design is to display all the
required information on a single screen, clearly and without distraction, in a
manner that can be assimilated quickly
Three layer of information
Monitoring
Analysis
Management
What to look for in a dashboard

Use of visual components to highlight data and exceptions that require
action.
Transparent to the user, meaning that they require minimal training and
are extremely easy to use
Combine data from a variety of systems into a single, summarized, unified
view of the business
Enable drill-down or drill-through to underlying data sources or reports
Present a dynamic, real-world view with timely data
Require little coding to implement/deploy/maintain
Best Practices in Dashboard Design
Benchmark KPIs with Industry Standards

Wrap the Metrics with Contextual Metadata
Validate the Design by a Usability Specialist
Prioritize and Rank Alerts and Exceptions
Enrich Dashboard with Business-User Comments
Present Information in Three Different Levels
Pick the Right Visual Constructs
Provide for Guided Analytics
Business Performance Management (BPM)
Business Performance Management (BPM) is…

 A real-time system that alerts managers to potential
opportunities, impending problems and threats, and then
empowers them to react through models and collaboration.
Also called corporate performance management (CPM by Gartner
Group), enterprise performance management (EPM by Oracle),
strategic enterprise management (SEM by SAP)
Business Performance Management (BPM)
BPM refers to the business processes, methodologies, metrics,

and technologies used by enterprises to measure, monitor, and
manage business performance.
BPM encompasses three key components
A set of integrated, closed-loop management and analytic processes,
supported by technology …
Tools for businesses to define strategic goals and then measure/manage
performance against them
Methods and tools for monitoring key performance indicators (KPIs), linked
to organizational strategy
Một số kỹ thuật cho mô hình dự báo
Một số kỹ thuật cho mô hình dự báo
1. Data Mining
2. Artificial neural networks
3. Support vector machines
4. The k-nearest neighbor
Data Mining Concepts/Definitions
Why Data Mining?
More intense competition at the global scale.

Recognition of the value in data sources.
Availability of quality data on customers, vendors, transactions,
Web, etc.
Consolidation and integration of data repositories into data
warehouses.
The exponential increase in data processing and storage
capabilities; and decrease in cost.
Movement toward conversion of information resources into
nonphysical form.
Definition of Data Mining
The nontrivial process of identifying valid, novel, potentially

useful, and ultimately understandable patterns in data stored in
structured databases. (Fayyad et al., (1996))
Keywords in this definition: Process, nontrivial, valid, novel,
potentially useful, understandable.
Data mining: a misnomer?
Other names: knowledge extraction, pattern analysis, knowledge
discovery, information harvesting, pattern searching, data
dredging,…
Data Mining là sự giao thoa của nhiều ngành khoa học
Ar
Pattern
tifi
cia
Recognition
l In
tic
tel
tis
lig
Sta
en
DATA
ce
Machine
MINING Learning
Mathematical
Modeling Databases
Management Science &

Information Systems
Data Mining Characteristics/Objectives
Source of data for DM is often a consolidated data warehouse

(not always!).
DM environment is usually a client-server or a Web-based
information systems architecture.
Data is the most critical ingredient for DM which may include
soft/unstructured data.
The miner is often an end user
Striking it rich requires creative thinking
Data mining tools’ capabilities and ease of use are essential
(Web, Parallel processing, etc.)
Data in Data Mining
Data: a collection of facts usually obtained as the result of

experiences, observations, or experiments.
Data may consist of numbers, words, images, …
Data: lowest level of abstraction (from which information and
knowledge are derived).
Data
Unstructured or
Structured
Semi-Structured
Categorical Numerical Textual Multimedia HTML/XML
Nominal Ordinal Interval Ratio
A Taxonomy for Data Mining Tasks
Data Mining Learning Method Popular Algorithms
Classification and Regression Trees,

Prediction Supervised
ANN, SVM, Genetic Algorithms
Decision trees, ANN/MLP, SVM, Rough

Classification Supervised
sets, Genetic Algorithms
Linear/Nonlinear Regression, Regression

Regression Supervised
trees, ANN/MLP, SVM
Association Unsupervised Apriory, OneR, ZeroR, Eclat
Link analysis Unsupervised Expectation Maximization, Apriory

Algorithm, Graph-based Matching
Sequence analysis Unsupervised Apriory Algorithm, FP-Growth technique
Clustering Unsupervised K-means, ANN/SOM
Outlier analysis Unsupervised K-means, Expectation Maximization (EM)
1.4 Quá trình khám phá tri thức
This is a view from typical machine learning and statistics

communities
Input Data Data Pre- Data Post-
Processing Mining Processing
Data integration Pattern discovery Pattern evaluation

Normalization Association & correlation Pattern selection
Feature selection Classification Pattern interpretation
Clustering
Dimension reduction Pattern visualization
Outlier analysis
…………
Data Mining Tasks (cont.)
Time-series forecasting
Part of sequence or link analysis?
Visualization
Another data mining task?
Types of DM
Hypothesis-driven data mining
Discovery-driven data mining
Data Mining Applications
Customer Relationship Management

Maximize return on marketing campaigns
Improve customer retention (churn analysis)
Maximize customer value (cross-, up-selling)
Identify and treat most valued customers
Banking & Other Financial

Automate the loan application process
Detecting fraudulent transactions
Maximize customer value (cross-, up-selling)
Optimizing cash reserves with forecasting
Data Mining Applications (cont.)
Retailing and Logistics

Optimize inventory levels at different locations
Improve the store layout and sales promotions
Optimize logistics by predicting seasonal effects
Minimize losses due to limited shelf life
Manufacturing and Maintenance

Predict/prevent machinery failures
Identify anomalies in production systems to optimize the use
manufacturing capacity
Discover novel patterns to improve product quality
Brokerage and Securities Trading

Predict changes on certain bond prices
Forecast the direction of stock fluctuations
Assess the effect of events on market movements
Identify and prevent fraudulent activities in trading
Insurance
Forecast claim costs for better business planning
Determine optimal rate plans
Optimize marketing to specific customers
Identify and prevent fraudulent claim activities
Computer hardware and software

Science and engineering
Government and defense
Homeland security and law enforcement
Travel industry
Healthcare Increasingly more
popular application areas
Medicine for data mining
Entertainment industry
Sports
Etc.
Data Mining Process
A manifestation of best practices

A systematic way to conduct DM projects
Different groups has different versions
Most common standard processes:
CRISP-DM (Cross-Industry Standard Process for Data Mining)
SEMMA (Sample, Explore, Modify, Model, and Assess)
KDD (Knowledge Discovery in Databases)
Data Mining Process: CRISP-DM
1 2
Business Data
Understanding Understanding
3
Data
Preparation
Data Sources
6
4
Deployment
Model
Building
5
Testing and
Evaluation
Neural Network Concepts
Neural networks (NN): a brain metaphor for information

processing
Neural computing
Artificial neural network (ANN)
Many uses for ANN for
pattern recognition, forecasting, prediction, and classification
Many application areas
finance, marketing, manufacturing, operations, information systems, and
so on
Biological Neural Networks
Two interconnected brain cells (neurons)
Dendrites
Synapse
Synapse
Axon
Axon
Dendrites Soma
Soma
Processing Information in ANN
A single neuron (processing element – PE) with inputs and

outputs
Inputs Weights Outputs
x1
w1 Y1
x2 w2 Neuron (or PE) f (S )

. S =
n
∑XW
Y
. Y2
. i i
.
.
i =1
Summation
Transfer
Function
.
wn Yn
xn
Biology Analogy
Elements of ANN
Processing element (PE)

Network architecture
Hidden layers
Parallel processing
Network information processing
Inputs
Outputs
Connection weights
Summation function
Elements of ANN
x1 (PE)
x2 Weighted Transfer
(PE) Sum Function
Y1
x3 (Σ) (f)
(PE)
(PE) (PE)
Output
(PE)
Layer
Hidden
(PE)
Layer Neural Network with
Input
One Hidden Layer
Layer
Elements of ANN
(a) Single neuron (b) Multiple neurons
x1 x1 w11 (PE) Y1
w1
w21
(PE) Y
w1 w12
x2 Y = X 1W1 + X 2W2
x2 w22 (PE) Y2
PE: Processing Element (or neuron)
Y1 = X 1W11 + X 2W21
w23
Summation Function for a Single Y2 = X 1W12 + X 2W22
Y3
Neuron (a), and Y3 = X 2W23 (PE)
Several Neurons (b)
Elements of ANN
Transformation (Transfer) Function

Linear function
Sigmoid (logical activation) function [0 1]
Tangent Hyperbolic function [-1 1]
Summation function: Y = 3(0.2) + 1(0.4) + 2(0.1) = 1.2

X1 = 3 Transfer function: YT = 1/(1 + e-1.2) = 0.77
W
1 =0
.2
W2 = 0.4 Processing Y = 1.2

X2 = 1
element (PE) YT = 0.77
.1
3
=0
W
X3 = 2
 Threshold value?
Neural Network Architectures
Architecture of a neural network is driven by the task it is

intended to address
Classification, regression, clustering, general optimization, association, ….
Most popular architecture: Feedforward, multi-layered perceptron
with backpropagation learning algorithm
Used for both classification and regression type problems
Others – Recurrent, self-organizing feature maps, Hopfield
networks, …
Feed-Forward Neural Networks
Feed-forward MLP with 1 Hidden Layer
Socio-demographic
= Predicted
vs. Actual
Religious
Voted “yes” or
“no” to legalizing
Financial gaming
. .
. .
. .
Other
INPUT HIDDEN OUTPUT

LAYER LAYER LAYER
Recurrent Neural Networks
Other Popular ANN Paradigms
Self-Organizing Maps (SOM)
Input 1  First introduced

by the Finnish
Professor Teuvo
Input 2
Kohonen
 Applies to
clustering type
problems
Input 3
Other Popular ANN Paradigms
Hopfield Networks
I n p u t
 First introduced
by John Hopfield
O  Highly
u
t interconnected
p
u
neurons
t  Applies to solving
complex
computational
problems (e.g.,
.
.
.
optimization
problems)
Development Process of an ANN
An MLP ANN Structure for the Box-Office Prediction
Problem
Class 1 - FLOP
1 (BO < 1 M)
MPAA Rating (5) Class 2

(G, PG, PG13, R, NR)
1 2 (1M < BO < 10M)
Competition (3) Class 3

(High, Medium, Low)
2 3 (10M < BO < 20M)
Star Value (3) Class 4

(High, Medium, Low)
3 4 (20M < BO < 40M)
Genre (10) Class 5

(Sci-Fi, Action, ... )
4 5 (40M < BO < 65M)
Technical Effects (3) Class 6

(High, Medium, Low)
5 6 (65M < BO < 100M)
Sequel (2) Class 7

6 7
(Yes, No)
... ... (100M < BO < 150M)
Number of Screens Class 8

(Positive Integer)
7 8 (150M < BO < 200M)
Class 9 - BLOCKBUSTER
9 (BO > 200M)
INPUT HIDDEN HIDDEN OUTPUT

LAYER LAYER I LAYER II LAYER
(27 PEs) (18 PEs) (16 PEs) (9 PEs)
Testing a Trained ANN Model
Data is split into three parts

Training (~60%)
Validation (~20%)
Testing (~20%)
k-fold cross validation

Less bias
Time consuming
AN Learning Process
A Supervised Learning Process
ANN
Model
 Three-step process:
 1. Compute temporary
Compute
output outputs.
 2. Compare outputs with
desired targets.
Adjust
weights
No Is desired
output
achieved?
 3. Adjust the weights and
repeat the process.
Yes
Stop
learning
Backpropagation Learning
Backpropagation of Error for a Single Neuron
α(Zi – Yi)
x1 error
w1
x2 w2 Neuron (or PE) f (S )

. S =
n
∑XW
Y = f (S ) Yi
. i i
.
i =1
Transfer
Summation
Function
wn
xn
Backpropagation Learning
The learning algorithm procedure

Initialize weights with random values and set other network parameters
Read in the inputs and the desired outputs
Compute the actual output (by working forward through the layers)
Compute the error (difference between the actual and desired output)
Change the weights by working backward through the hidden layers
Repeat steps 2-5 until weights stabilize
Support Vector Machines (SVM)
Goal of SVM: to generate mathematical functions that map input

variables to desired outputs for classification or regression type
prediction problems.
First, SVM uses nonlinear kernel functions to transform non-linear
relationships among the variables into linearly separable feature spaces.
Then, the maximum-margin hyperplanes are constructed to optimally
separate different classes from each other based on the training dataset.
SVM has solid mathematical foundation!
Support Vector Machines (SVM)
A hyperplane is a geometric concept used to describe the

separation surface between different classes of things.
In SVM, two parallel hyperplanes are constructed on each side of the
separation space with the aim of maximizing the distance between them.
A kernel function in SVM uses the kernel trick (a method for
using a linear classifier algorithm to solve a nonlinear problem)
The most commonly used kernel function is the radial basis function (RBF).
Application Case 6.4
Managing Student Retention with Predictive Modeling
Questions for Discussion

Why is attrition one of the most important issues in higher
education?
How can predictive analytics (ANN, SVM, and so forth) be used
to better manage student retention?
What are the main challenges and potential solutions to the use
of analytics in retention management?
Application
Case 6.4
Managing Student Retention with Predictive Modeling
How Does an SVM Work?
Following a machine-learning process, an SVM learns from the

historic cases.
The Process of Building SVM
1. Preprocess the data
Scrub and transform the data.
2. Develop the model.
Select the kernel type (RBF is often a natural choice).
Determine the kernel parameters for the selected kernel type.
If the results are satisfactory, finalize the model; otherwise change the kernel type and/or kernel parameters to
achieve the desired accuracy level.
3. Extract and deploy the model.
The Process of Building an SVM
Pre-Process the Data

Training
ü Scrub the data
data
“Identify and handle missing,
incorrect, and noisy”
ü Transform the data
“Numerisize, normalize and
standardize the data”
Pre-processed data
Develop the Model

Experimentation
ü Select the kernel type “Training/Testing”
“Choose from RBF, Sigmoid
or Polynomial kernel types”
ü Determine the kernel values
“Use v-fold cross validation or
employ ‘grid-search’”
Validated SVM model
Deploy the Model

Prediction
ü Extract the model coefficients Model
ü Code the trained model into
the decision support system
ü Monitor and maintain the
model
SVM Applications
SVMs are the most widely used kernel-learning algorithms for

wide range of classification and regression problems
SVMs represent the state-of-the-art by virtue of their excellent
generalization performance, superior prediction power, ease of
use, and rigorous theoretical foundation
Most comparative studies show its superiority in both regression
and classification type prediction problems.
SVM versus ANN?
k-Nearest Neighbor Method (k-NN)
ANNs and SVMs  time-demanding, computationally intensive

iterative derivations
k-NN is a simplistic and logical prediction method, that produces
very competitive results
k-NN is a prediction method for classification as well as
regression types (similar to ANN & SVM)
k-NN is a type of instance-based learning (or lazy learning) –
most of the work takes place at the time of prediction (not at
modeling)
k : the number of neighbors used
k-Nearest Neighbor Method (k-NN)
k=3
k=5
Yi
The answer depends on

the value of k
Xi X
The Process of k-NN Method
Training Set
Parameter Setting
Historic Data ü Distance measure

ü Value of “k”
Validation Set
Predicting
Classify (or Forecast)
new cases using k
number of most
similar cases
New Data
k-NN Model Parameter
Similarity Measure: The Distance Metric
Numeric versus nominal values?
k-NN Model Parameter
Number of Neighbors (the value of k)

The best value depends on the data
Larger values reduce the effect of noise but also make boundaries between
classes less distinct
An “optimal” value can be found heuristically
Cross Validation is often used to determine the best value for k
and the distance measure
Efficient Image Recognition and Categorization with kNN
Questions for Discussion

Why is image recognition/classification a worthy but difficult
problem?
How can k-NN be effectively used for image
recognition/classification applications?
Một số kỹ thuật ra quyết định dựa
trên mô hình
Major Modeling Issues
Problem identification and environmental analysis (information

collection)
Variable identification
Influence diagrams, cognitive maps
Forecasting/predicting
More information leads to better prediction
Multiple models: An MSS can include several models, each of
which represents a different part of the decision-making problem
Categories of models >>>
Model management – DBMS vs. MBDM
Categories of Models
Model Categories
Static and Dynamic Models
Static Analysis
Single snapshot of the situation
Single interval
Steady state
Dynamic Analysis
Dynamic models
Evaluate scenarios that change over time
Time dependent
Represents trends and patterns over time
More realistic: Extends static models
Model Categories
Current Trends in Modeling
Development of Model/Solution Libraries

NEOS Server for Optimization
 neos.mcs.anl.gov/neos/index.html
Resources link at informs.org
 lionhrtpub.com/ORMS.shtml
Web-based modeling (optimization/simulation/…)
Multidimensional analysis (modeling)
Influence Diagrams
Examples - Components of Models
The Structure of a
Mathematical Model
 The components of a quantitative model are linked

together by mathematical (algebraic) expressions—
equations or inequalities.
 Example – Profit - 𝑃𝑃 = 𝑅𝑅 − 𝐶𝐶
whereP= profit, R= revenue, and C= cost

 Example - Simple Present-Value -
𝐹𝐹 100,000
𝑃𝑃 = 𝑛𝑛
= 5
= 62,092
(1 + 𝑖𝑖) 1 + 0.1
whereP= present value, F= future cash-flow, i=
interest-rate, and n = number of period (years)
Modeling and Decision Making -
Under Certainty, Uncertainty, and Risk
Certainty
Assume complete knowledge
All potential outcomes are known
May yield optimal solution
Uncertainty
Several outcomes for each decision
Probability of each outcome is unknown
Knowledge would lead to less uncertainty
Risk analysis (probabilistic decision making)
Probability of each of several outcomes occurring
Level of uncertainty => Risk (expected value)
Modeling and Decision Making -
Under Certainty, Uncertainty, and Risk
The Zones of Decision Making
Decision Modeling with Spreadsheets
Spreadsheet
Most popular end-user modeling tool
Flexible and easy to use
Powerful functions (add-in functions)
Programmability (via macros)
What-if analysis and goal seeking
Simple database management
Seamless integration of model and data
Incorporates both static and dynamic models
Examples: Microsoft Excel, Lotus 1-2-3
Optimization
via Mathematical Programming
Mathematical Programming
 A family of tools designed to help solve managerial problems in which the
decision maker must allocate scarce resources among competing activities
to optimize a measurable goal
Optimal solution: The best possible solution to a modeled

problem
Linear programming (LP): A mathematical model for the optimal solution
of resource allocation problems. All the relationships are linear.
Spreadsheet Model Helps Assign Medical Residents

Company
Problem description
Proposed solution
Results
LP Problem Characteristics
Limited quantity of economic resources

Resources are used in the production of products or services
Two or more ways (solutions, programs) to use the resources
Each activity (product or service) yields a return in terms of the
goal
Allocation is usually restricted by constraints
Linear Programming Steps
Identify the …
Decision variables
Objective function
Objective function coefficients
Constraints
Capacities / Demands / …
Represent the model

LINDO: Write mathematical formulation
EXCEL: Input data into specific cells in Excel
Run the model and observe the results
Modeling in LP - An Example
The Product-Mix Linear Programming Model

MBI Corporation
Decision variable: How many computers to build next month?
Two types of mainframe computers: CC-7 and CC-8
Constraints: Labor limits, Materials limit, Marketing lower limits
CC-7 CC-8 Rel Limit
Labor (days) 300 500 <= 200,000 /mo
Materials ($) 10,000 15,000 <= 8,000,000 /mo
Units 1 >= 100
Units 1 >= 200
Profit ($) 8,000 12,000 Max
Objective: Maximize Total Profit / Month
Common Optimization Models
Product-mix problems (how many of each product to produce for

max profit)
Transportation (minimize cost of shipments)
Assignment (best matching of objects)
Investment (maximizing rate of return)
Network optimization models for planning and scheduling
Replacement (capital budgeting), …
Multi-Criteria Decision Making with Pairwise
Comparisons
Having more than one criterion makes decision-making process

complicated
Usually some type of weighing algorithm is used to analyze such
problems
The Analytic Hierarchy Process
Developed by Thomas Saaty (1995, 1996)
A very popular technique for MCDM
Popular Tools - ExpertChoice.com
Web-based Tools - Web-HIPRE (hipre.aalto.fi)
Hệ ra quyết định tự động và Hệ
chuyên gia
Automated Decision Systems
A relatively new approach to supporting decision making

a.k.a. Decision Automation Systems (DAS)
Often a rule-based system that provides a solution in a functional
area
“If only 70 percent of the seats on a flight from LA to NY are sold 3 days
prior to departure, offer a discount of x to nonbusiness travelers”
Applies to repetitive/structured decisions
Automated Decision-Making Framework
Architecture of the Airline Revenue Management
Systems
Artificial Intelligence (AI)
Artificial intelligence (AI)

A subfield of computer science, concerned with symbolic reasoning and
problem solving
AI has many definitions…

Behavior by a machine that, if performed by a human being, would be
considered intelligent
“…study of how to make computers do things at which, at the moment,
people are better
Theory of how the human mind works
AI Objectives
Make machines smarter (primary goal)

Understand what intelligence is
Make machines more intelligent & useful
Signs of intelligence…
Learn or understand from experience
Make sense out of ambiguous situations
Respond quickly to new situations
Use reasoning to solve problems
Apply knowledge to manipulate the environment
Test for Intelligence
Turing Test for

Intelligence
A computer can be
considered to be smart
only when a human
interviewer, “conversing” Questions / Answers
with both an unseen

human being and an
unseen computer, can not
determine which is which.
 - Alan Turing
Expert Systems (ES)
Is a computer program that attempts to imitate expert’s

reasoning processes and knowledge in solving specific problems
Most Popular Applied AI Technology
Enhance Productivity
Augment Work Forces
Works best with narrow problem areas/tasks
Expert systems do not replace experts, but
Make their knowledge and experience more widely available, and thus
Permit non-experts to work better
Important Concepts in ES
Expert
 A human being who has developed a high level of proficiency in
making judgments in a specific domain
Expertise
The set of capabilities that underlines the performance of human experts,
including
extensive domain knowledge,
heuristic rules that simplify and improve approaches to problem solving,
meta-knowledge and meta-cognition, and
compiled forms of behavior that afford great economy in a skilled performance
Features and Concepts in ES
Experts / Expertise
Degrees or levels of expertise
Ratio of non-experts to experts  100 to 1
Transferring Expertise
From expert to computer to nonexperts via acquisition, representation,
inferencing, transfer
Symbolic Reasoning / Inferencing
Deep Knowledge / Self Knowledge
Applications of Expert Systems
Classical Applications
DENDRAL
Applied knowledge (i.e., rule-based reasoning)
Deduced likely molecular structure of compounds
MYCIN
A rule-based expert system
Used for diagnosing and treating bacterial infections
XCON
A rule-based expert system
Used to determine the optimal information systems configuration
New applications: Credit analysis, Marketing, Finance,

Manufacturing, Human resources, Science and Engineering,
Education, …
Applications of Expert Systems
Structure of Expert Systems
Development Environment
Consultation Environment
Major Components
Knowledge acquisition subsystem
Knowledge Engineer
Knowledge Base
Inference Engine
User Interface
Blackboard (workplace)
Explanation subsystem (justifier)
Knowledge-refining system
Structures of Expert
Systems
en nt
nm e
t
ro pm
Human
vi lo
Other Knowledge
En eve
Expert(s)
Sources
D
en n
nm tio
Knowledge Information
t
ro ta
Elicitation Gathering
vi sul
En on
C
Knowledge
Rules
Knowledge
Knowledge Base(s)
Engineer (Long Term)
Inferencing
Rules
Rule
Questions Inference Engine Firings
/ Answers
Explanation Knowledge
User Facility Refinement Refined
User Rules
Interface
Blackboard (Workspace)
Facts Data /
Facts Information
Working External Data

Memory Sources
(Short Term) (via WWW)
Nguồn tri thức
Chuyên gia
khác
Cơ sở tri thức
(dài hạn)
Hỏi/Đáp
Giao tiếp
gười dùng
Sự kiện Sự
Vùng nhớ
làm việc
(ngắn hạn)
This Photo by Unknown Author is licensed
under CC BY-SA

7

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

7

Uploaded by

Copyright:

Available Formats

KHO DỮ LIỆU VÀ HỆ THỐNG HỖ

TRỢ RA QUYẾT ĐỊNH

1. Hiểu về báo cáo kinh doanh

1. Báo cáo kinh doanh

Báo cáo = Thông tin  Quyết định

Rõ ràng (Clarity) …

Các đặc tính chung

Data visualization can date back to the second century AD

• Arguably the most popular multi-dimensional chart By Charles Joseph Minard

Which one to use? Where and when?

Các đơn vị hàng đầu

Magic Quadrant for Business Intelligence and Analytics Platforms

Source: Gartner (February 2020)

Emergence of new companies

A recently coined term

SAS Visual Analytics Architecture

At teradatauniversitynetwork.com, you can learn more about SAS

Performance dashboards are commonly used in BPM software

What to look for in a dashboard

Benchmark KPIs with Industry Standards

Business Performance Management (BPM) is…

BPM refers to the business processes, methodologies, metrics,

More intense competition at the global scale.

The nontrivial process of identifying valid, novel, potentially

Management Science &

Source of data for DM is often a consolidated data warehouse

Data: a collection of facts usually obtained as the result of

Categorical Numerical Textual Multimedia HTML/XML

Nominal Ordinal Interval Ratio

Data Mining Learning Method Popular Algorithms

Classification and Regression Trees,

Decision trees, ANN/MLP, SVM, Rough

Linear/Nonlinear Regression, Regression

Association Unsupervised Apriory, OneR, ZeroR, Eclat

Link analysis Unsupervised Expectation Maximization, Apriory

Sequence analysis Unsupervised Apriory Algorithm, FP-Growth technique

Clustering Unsupervised K-means, ANN/SOM

Outlier analysis Unsupervised K-means, Expectation Maximization (EM)

This is a view from typical machine learning and statistics

Data integration Pattern discovery Pattern evaluation

Customer Relationship Management

Banking & Other Financial

Retailing and Logistics

Manufacturing and Maintenance

Brokerage and Securities Trading

Computer hardware and software

A manifestation of best practices

Neural networks (NN): a brain metaphor for information

Two interconnected brain cells (neurons)

A single neuron (processing element – PE) with inputs and

x2 w2 Neuron (or PE) f (S )

Processing element (PE)

(a) Single neuron (b) Multiple neurons

Several Neurons (b)

Transformation (Transfer) Function

Summation function: Y = 3(0.2) + 1(0.4) + 2(0.1) = 1.2

W2 = 0.4 Processing Y = 1.2

Architecture of a neural network is driven by the task it is

Feed-forward MLP with 1 Hidden Layer

INPUT HIDDEN OUTPUT

Input 1  First introduced

MPAA Rating (5) Class 2

Competition (3) Class 3

Star Value (3) Class 4