Please enable JavaScript.

Coggle requires JavaScript to display documents.

MODULE 4 DATA ANALYTICS, INTERPRETATION AND VISUALISATION - Coggle Diagram

- - - - Making decisions based on knowledge created through analytics
      - Decision maker combines knowledge from data analysis with their own knowledge to make decisions
      - Not data-dictated (bắt buộc) decisions
    - - Definition:
        
        Data: little use in raw state
        
        Analytics: derive information from data using of skills, processes and technologies
        
        Information: meaning and purpose relevant to needs
        
        Knowledge: integrating info with prior exp, skills and opinion
        
        Wisdom: use knowledge to make decision
      - 4 basics forms of analytics (p.245)
        
        Descriptive analytics
        
        Diagnostic analytics
        
        Predictive analytics
        
        Prescriptive analytics
      - Data Science
        
        a set of fundamental principles that guide the creation of knowledge from data
        
        4 Key principles
        
        Using mathematics, statistics and probabilities to extract knowledge from data
        
        Using a systematic process to extract that knowledge
        
        Fitting models to data
        
        Avoiding patterns that do not generalize beyond the data set
      - Data Mining
        
        the use of technology based on data science
        
        Technologies applied
        
        AI
        
        ML
        
        Data visualization tools
    - - 90% of strategies refer analytics as essentials
      - Only 20% insights convert into outcomes
  - - - Definition
        
        Data Literacy = ability to read, work with, analyse and argue with data
      - Details:
        
        Read: ability to look and comprehend data
        
        Work: how to create, acquire, clean and manage data
        
        Analyse: draw insights from the data by applying analytical techniques
        
        Argue: use data to support idea, action or theory
    - - Data is important assets
      - Data literacy helps
        
        Organisation
        
        Improve efficiency and effectiveness of products/services
        
        Maximise the value from data assets
        
        Employees
        
        Increase productive in their daily tasks
        
        Make data-informed decisions
      - require understanding some
        basic maths, statistics and the fundamental properties of data
    - - using analytics to reduce operational and reporting costs
      - encouraging the use of data to inform decision making
      - promoting the ‘democratisation of analytics’
      - sharing data and insights with external stakeholders
  - - - Definition: organisational behaviour that embrace the use of
        data-informed decision making and foster data literacy
      - dựa trên sự thẩm thấu trong toàn tổ chức, starting with adoption from the organisation's leaders to a top-down approach
      - phân loại
        người dùng thành
        
        Data personas
        (level a literacy of a person has)
        
        Data Aristocrat
        
        Data Knight
        
        Data Dreamer
        
        Data Doubter
    - - From least to most data literate
        
        Data Doubter
        
        Data Dreamer
        
        Data Knight
        
        Data Aristocrat
        
        Roles: learning new analytics algorithms, methodologies, and tools, and developing their skills in leadership and mentoring
        
        high level of skill and experience in data analytics
        
        in the process of building their data science skills and knowledge
        
        Roles: develop their use of data in their work, and their skills in data science
        
        basic awareness of data and desire to learn
        
        Roles: continue to learn and develop their skills
        
        sceptical of the value of data and data-informed decision making
        
        fear the complexity of the data analytics
        
        Roles: Other personas should communicate and demonstrate the benefits of data analytics to Data Doubter
      - Data Aristocrat and Data Knight = Data Champions
      - Senior leadership with Data Aristocrat = Data Leaders
  - - - Qualitative
        (Categorical)
        
        Dữ liệu định tính:
        chứa các yếu tố không thể đo lường
        
        Nominal
        
        cannot be ordered
        from low to high
        
        Ordinal
        
        can be ranked
        from low to high
      - Quantitative (Numberical)
        
        Dữ liệu định lượng:
        chứa các yếu tố có thể đo lường
        
        Interval
        
        = ordinal + no natural zero starting point
        
        Phân nhóm 2
        
        Discrete
        
        chỉ có số nguyên (integer number)
        
        Continuos
        
        bao gồm cả số thực (real number)
        
        Ratio
        
        = interval + had a natural zero starting point
    - - the chance that an event will occur
      - Probability Theory terms and concepts
        
        Types of Probability
        
        Objective Probability
        Xác suất khách quan
        (based on real analysis)
        
        Priori probability Xác suất tiên nghiệm (dựa trên kiến thức và suy luận logic)
        
        Emperical probability Xác suất thực nghiệm (dựa trên dữ liệu quan sát và thực nghiệm)
        
        Subjective Probability
        Xác suất chủ quan
        (based on experience)
        
        Unconditional Probability
        Xác suất không điều kiện
        [P(A) và P(B)]
        
        Conditional Probability
        Xác suất có điều kiện
        [P(A|B)] = P(A) / P(B)
        
        Total Probability
        
        Outcomes: single result
        
        Events:
        set of outcomes
        
        Dependent Events: Biến cố phụ thuộc
        [P(A|B)]
        
        Independent Events: Biến cố độc lập
        P(A) hoặc P(B)
        
        Mutually Events: Biến cố xung khắc
        [P (A ∩ B) = 0]
        
        Collectively Exhaustive Events: Biến cố đầy đủ
        [P (A ∪ B)]
        
        Bayes' Theorem
        
        P(A|B) = [P(B|A)] x P(A) / P(B)
        
        Partition: a collection of outcome sets (không overlap)
    - - Structured data - Dễ dàng thu thập, truy xuất, sắp xếp, lưu trữ và trích xuất/ Định dạng: spreadsheet hoặc SQL
      - Semi-structured data - Chứa cấu pháp đơn giản và cần mapping, processing trước khi sử dụng / Định dạng: XML, CSV
      - Unstructured data - Khó thu thập, truy xuất, lưu trữ và sắp xếp/ Định dạng: photo, ghi âm, etc.
- - - - Business strategy
        
        strategy priorities
        
        info needs
        
        use cases
        
        Data requirement
        
        determined by info needs
        
        how to source
        
        Analytics requirement
        
        choose options
        
        ensure invest in analytics
        
        Communication
        
        clear communication of outputs and insights
        
        Interpretation and application
        
        decision making
        
        action
        
        Implementation
        
        data governance
        
        systems
        
        change management
  - - - Defence-offence
        framework
        
        trade-offs between
        
        risk mitigation and benefits of data usage
        
        control and flexibility
      - Tableau framework
        
        tích hợp business với IT strategy
        
        identify 3 steps
        
        Define business objectives
        
        Capture the right data to support objectives
        
        Mordernize the data architecture to support data volume, variety and velocity
      - SAS data strategy framework
        
        focus on data management
        
        identify 5 components
        
        Identify
        
        Store
        
        Provision
        
        Process
        
        Govern
      - Gartner approach
        
        focuses on how data and analytics technical professionals can create a data strategy
        
        8-step process
      - SMART strategy board
        
        SMART stands for
        
        Strategy
        
        Measure metric and data
        
        Apply analytics
        
        Report the result
        
        Transform the business
        
        SMART strategy board
        
        Purpose panel
        
        reflect business strategy by details:
        
        Mission statement
        
        Vision statement
        
        Customer panel
        
        shows how much its know about customers by consider:
        
        Target market
        
        Value proposition
        
        Finance panel
        
        prompts how much its deliver financial result and it still need to learn
        
        Operation panel
        
        consider what its need to do internally
        
        Partners
        
        Core competence
        
        Finance panel
        
        consider info and resources need to achieve strategic objectives
        
        IT resources
        
        Infrastructure
        
        Human resource
        
        Value and leadership
        
        Competition and risk panel
        
        consider the competitors and risks must be managed
        
        Competition
        
        Risk
        
        Competition and risk data
        
        Fraud data
        
        Search engine data
        
        Social media data
      - help to identify trends in data and make business decisions
    - - 7 ways to generate data
        
        Transaction data
        
        generated when customer buys something
        
        Internal - Structured
        
        Created data
        
        when people are asked questions and their answers
        
        Internal/External - Structured/Semi-structured
        
        Provoked data
        
        when people are invited to express their views
        
        Internal/External - Structured/Semi-structured
        
        Complied data
        
        from the giant databases built by research and analysis companies
        
        External - Structured
        
        Experimental data
        
        a hybrid of created and transaction data
        
        Internal/External - Structured/Semi-structured
        
        Captured data
        
        gathered passively from an individual’s behaviour
        
        Internal/External Unstructured
        
        User-generated data
        
        individuals and companies produce consciously, or at least knowingly
        
        Internal/External Unstructured
    - - require the application of AI
        
        Diagnostic analytics
        determine cause and effect
        
        Descriptive analytics
        summarise performance
      - conventional statistical methods
        
        Predictive analytics
        forecast future performance
        
        Predictive analytics
        forecast future performance
  - - - clearly define business strategy
      - identify targets from strategy
      - identify KPIs
      - ranking the priorities of KPIs
      - create a plan to achieve goals and obj
      - estimate the demand of costs for implementation
      - update the priorities based on results and changes
    - - Often everyone argees to establish KPIs
        but don't check if the quality of data
        is good enough to measure KPIs
        
        Guidelines to fix the issue
        (Result: Build data culture)
        
        Make data and analytics
        accessible and pervasive
        
        Send message from the top down
        
        Connect data to performance
        
        Show why it work
        
        Acknowledge uncertainty
      - Definition: Data ownership -
        who is responsible and
        accountable for managing data
    - - Data managment
        
        the process that oversees the data life cycle
        
        Objectives:
        
        Reduce risks & costs of complying with regulations
        
        Follow legal requirements
        
        Safeguarding data security
        
        Maintaining accuracy
        
        Certifying consistency
        
        Access, storage,
        backup and disposal
      - Data life cycle
        
        link between data sources
        to business applications
        
        5 phases
        
        Databases
        
        Data storage
        
        Data analysis
        
        Results
        
        1 more item...
        
        Higher level
        
        System Development Life Cycle
        (SDLC)
        
        Planning
        Create a plan
        
        Analysis
        Planning how and how much to do it
        
        Design
        Create system design specification
        
        1 more item...
      - Database management system (DBMSs)
        
        DBMS functions
        
        Data integrity
        and maintenance
        
        Data synchronisation
        
        Data security
        
        Data access
        
        Data filtering
        and profiling
        
        How data quality decrease
        
        incomplete
        
        outdated
        
        incorrect
        
        duplicate
        
        non-standarlised
        
        unusable
        
        Cost of poor quality data
        = Lost business + Cost to prevent errors + Cost to correct errors
    - - migrate data from old system to new systems
      - Data Migration
        Process
        
        Data mapping is
        first step and a critical step
        
        Data cleansing -
        important to improve data quality
    - - Kotter's 8-step Framework
        
        Step 1: Establish sense of urgency
        
        Step 2: Create a guideline
        
        Step 3: Develop a vision and strategy
        
        Step 4: Communicate the vision
        
        Step 5: Empower other people to change
        
        Step 6: Generate short-term wins
        
        Step 7: Consolidate gains and produce more changes
        
        Step 8: Anchor new approaches in the culture
        
        need to perform well in system transformation
      - manage the change of systems in order to success
- - - - Definition: data sets
        cannot be stored or processed
        by traditional data tools and methods
      - Where to generate:
        
        Business
        
        Gov' agencies
        
        Financial market
        
        Sensors
        
        IoT
      - 4 V
        (module 1)
        
        Volume
        
        Variety
        
        Velocity
        
        Veracity
        
        5th 'V' - Value
      - Types of analytics
        
        Conventional
        (Normal) Analytics
        
        Phân tích chọn mẫu để đưa ra cho toàn bộ
        
        Regression model
        
        describe and quantify ‘on average’ relationships
        
        predict new behaviours
        
        forecast future values
        
        Big Data Analytics
        
        deal large data sets and mix of structured and unstructured
        
        outcomes of data mining
        
        find patterns, shared characteristics and classify
        
        model relationship b/w variables
        
        find outliers and noise
    - - Reasons for adopting big data
        
        Consolidation - Tích hợp lưu trữ dữ liệu vào 1 nguồn duy nhất
        
        Data-driven insight
        
        Analytics workforce - Tận dụng nguồn lực để phân tích và diễn giải thay vì chỉ thu thập dữ liệu
        
        Visualisation - Trình bày dữ liệu dễ hiểu, dễ tương tác
        
        Operational efficiency - Faster decision making
        
        Enhanced productivity
        
        Cost reduction
        
        Competitive advantage - New products/ services
    - - Elements before implementation
        
        Assess and strategies - Evaluate current situation and determine a suitable strategy
        
        Stakeholder involvement - Explain vision and mission to stakeholder
        
        Create a clear plan - Create a strategic plan
      - 5 khía cạnh của dữ liệu when implementation
        
        Ingestion (Nhập) - Quá trình lấy và nhập dữ liệu để sử dụng và lưu trữ
        
        Harmonisation (Hoà hợp) - Hợp nhất dữ liệu từ nhiều nguồn thành 1 tệp dữ liệu gắn kết duy nhất
        
        Analysis - Cleaning, transforming, and modelling data
        
        Visualisation - Trình bày dạng biểu đồ
        
        Democratisation (Dân chủ hoá) - making digital information accessible to a non-technical user of information systems
      - Challenges
        
        Access to good data
        
        Ability to analyse
        
        Ability to turn analysis into insights
        
        Dealing unstructured data
        
        High costs
        
        Data privacy issues
        
        Weak data visualisation
        
        Poor governance and management
        
        Get the right analytics skills
        
        Inadequate infrastructure
        
        Cultural change
    - - New Skills
        
        Xây dựng strong understanding of digital tech on finance
        
        Trình bày giá trị của digital processes and systems
        
        Phát triển và thông thạo
        nền tảng kỹ thuật số
        
        Khen thưởng với curiosity about digital applications
      - Core Skills
        
        Data communication and visualisation skills
        
        Các công cụ và ngôn ngữ lập trình cho phép người dùng truy cập, truy xuất, truy vấn và trình bày dữ liệu
        
        Xử lý dữ liệu
        
        Kiến thức về các phương pháp thống kê và khi nào sử dụng chúng
      - Developing Skills
        
        Grasp unfamiliar data
        
        Invest sufficient time
        
        Develop insights
        
        Build context through additional research
        
        Communicate clearly
        
        Ask questions
  - - - Step 1: Define purpose
        
        Step 2: Obtain data
        
        Step 3: Explore & clean data
        
        Step 4: Reduce data dimension (loại bỏ dữ liệu thừa)
        
        Step 5: Determine analysis task
        
        Step 6: Partition data
        (Phân vùng dữ liệu theo purpose)
        
        1 more item...
    - - Data Types
        
        Structured
        
        Semi-structured
        
        Unstructured
      - Data Sources
        
        Internal
        
        External
      - Data Ethics and Regulation
        
        Informed Consent
        
        Anonymising Data
        
        Biased Data and Biased Analytics
    - - Training partition data
      - Validation set
      - Test data set
    - - Testing and evaluating models
        
        More variables include, more risk of overfitting
      - Caveats and other principles
        
        Symmetry
        
        Occam's razor
        
        No free lunch theorem
        
        Ensembles
  - - - Require accuracy in data classification => greater quality of model applied
    - - Greater volume of data => greater encountering error values
    - - 3 cases
        
        Chiếm small proportion
        
        Có thể bỏ qua
        
        Chiếm considerable proportion
        
        Thay thế missing values bằng imputed values để giảm sai số với các giá trị còn lại
        
        Chiếm large proportion
        
        Examine the importance of predictor
        
        Yes
        
        Invest to obtain the missing data from other sources
        
        No
        
        Dừng/ Drop
    - - Some algorithms required data
        to be standardising to bring
        all variables to same scale
    - - Cost of Misclassification
      - Failure to locate responders
    - - Integrating climate risk => better plan for future disasters
  - - - Data Wrangling (Sắp xếp data)
        
        quá trình chuyển đổi dữ liệu
        từ dạng thô sang usable format
        
        ML techniques applied
        
        remove duplicate records
        
        standardise value formats
        
        validate the accuracy of data
        
        generate new data from the existing clean data
        
        automate some tasks
      - Analysing
        unstructed data
        
        AI techniques such as NLP can extract data from these media formats
      - Finding
        Unknown patterns
        
        Unsupervised ML allows the discovery of patterns, trends, collections and correlations in the data
      - Real-time
        monitoring
        
        AI can monitor continuous, real-time events
        quickly detect and highlight anomalous events
      - Automation
        Action
        
        AI can automate actions based on the results of analytic processes.
      - Natural Language
        Analytics
        
        NLP to allow users to conduct queries in intuitive language
    - - Poor
        Data quality
        
        AI techniques depend on high-quality data
      - A lack of Algorithmic transparency
        
        Hidden workings of the AI-derived model mean that any biases introduced by the data are unclear and the rationale for the output cannot be examined and validated
    - - Supervised learning
        
        Definition: thuật toán dự đoán đầu ra (outcome) của một dữ liệu mới (new input) dựa trên các cặp (input, outcome) đã biết từ trước
        
        Input data + Output data tương ứng (label) = Algorithms
        thể hiện quan hệ giữa input data và output data
        
        2 Supervised learning
        techniques
        
        Classification
        
        Nếu các label của input data được chia thành một số hữu hạn nhóm
        
        Regression
        
        Nếu label không được chia thành các nhóm mà
        xác định variables quan trọng ảnh hưởng đến outcome
      - Unsupervised learning
        
        Only have input data, no label
        
        Sử dụng phân nhóm (clustering) hoặc giảm số chiều của dữ liệu (dimension reduction)
        để thuận tiện trong việc lưu trữ và tính toán.
        
        From Input data => tự tìm ra algorithms
        
        Unsupervised learning techniques
        
        Clustering
        
        Phân nhóm input data thành các nhóm nhỏ dựa trên sự liên quan giữa các dữ liệu của nhóm
        
        Association rule mining
        
        Khám phá ra một quy luật dựa trên nhiều dữ liệu cho trước
        
        Collaborative filtering
        
        predict users' interests and recommend items that would appeal to them.
        
        Approaches
        
        1st approach
        
        based around properties of an item
        
        2nd approach
        
        similarities in preferences between users
      - Semi-supervised learning
        
        Definition: thuật toán có một lượng lớn dữ liệu nhưng chỉ một phần đã gắn nhãn (label)
        
        Giai đoạn đầu: dùng cặp data- label có sẵn để tạo algorithms
        Giai đoạn sau: sử dụng algorithms tạo thành để đào tạo unlabeled data ra các predicted outcome
        rồi sử dụng cặp mới tạo thành (data - pseudo-labelled) tiếp tục update algorithms
      - Dimensionality Reduction for ML
        
        giảm số lượng các input không cần thiết trong một tập dữ liệu
      - Useful ML Algorithms
        
        Decision trees
        
        Used both in supervised learning and unsupervised learning
        
        nonlinear and used for categorical or discrete values
        
        Support Vector Machines
        
        Used in supervised learning, mostly classification
        
        Support Vector Machine là bài toán đi tìm mặt phân cách sao cho margin tìm được là lớn nhất, đồng nghĩa với việc các điểm dữ liệu an toàn nhất so với mặt phân cách.
- - - - DIWK Pyramid
        
        Data
        
        Information
        
        Knowledge
        
        Wisdom
        
        Knowledge + insights
        
        Information + meaning
        
        Data + context (bối cảnh)
        
        Dữ liệu thô
        
        Interpretation = giai đoạn chuyển từ Information sang Knowledge
    - - kiến thức về một chuyên ngành hoặc
        lĩnh vực cụ thể, chuyên biệt,
        different với general knowledge
      - e.g: software developer (general knowledge)
        có kinh nghiệm ở bank sector (domain knowledge)
    - - Quantitative analysis
        
        describes the use of mathematical and statistical techniques
        to analyse numerical data
      - Qualitative analysis
        
        relies on critical thinking and the subjective interpretation of data
        
        Input: unstructured data w/t qualitative
        
        3 Steps
        
        Coding
        
        Analysis
        
        Summarising
    - - 2 key roles in analytics
        
        giải thích dữ liệu và kết quả phân tích trong và sau quá trình phân tích
        
        trình bày các phát hiện và khuyến nghị
        trong quá trình báo cáo
      - Sample approaches
        
        Magnitude comparison
        
        Change over time
        
        Correlation
    - - Cognition bias
        (Bias trong tư duy)
        
        prejudice influences
        our decisions or thinking
      - Types
        
        Confirmation bias
        
        khuynh hướng đi tìm những thông tin xác nhận
        và củng cố những định kiến bản thân về một vấn đề
        
        Anchoring
        
        khuynh hướng quá tin với điều đầu tiên bạn nghe về một chủ đề
        
        Herd Behaviour
        
        khuynh hướng chấp nhận niềm tin hoặc
        giá trị được số đông người khác cùng tin
        
        Optimism Bias
        
        khuynh hướng lạc quan thái quá,
        phóng đại xác suất thành công và xem nhẹ rủi ro
        
        Extrapolation Bias
        
        khuynh hướng xem một số sự kiện dễ dự đoán
        dựa vào simple analysis, dù nhiều variables liên quan
        
        Availability Bias
        
        khuynh hướng tin rằng thông tin bạn sở hữu quan trọng hơn thực tế
      - Strategies to overcome cognition bias
        
        Awareness
        
        Collaboration
        
        Alternative viewpoints
    - - Definition:
        mối quan hệ giữa hai variables
      - Types
        
        Positive
        correlation
        
        2 variables move cùng direction
        
        Negative
        correlation
        
        2 variables move opposite direction
      - Correlation coefficient
        
        measure a strength of correlation
        
        Formula
        
        Cov(x, y)
        −−−−−−−
        (σx * σy)
    - - Definition:
        mối quan hệ nhân quả
        giữa hai variables
      - Nếu 2 variables không có mối quan hệ nhân quả thì sẽ là correlation
        => Spurious association
      - Chỉ tồn tại nếu 1 variable gây ảnh hưởng trực tiếp đến variable còn lại
      - Proving causation
        
        Random Experiments
        
        Chỉ định một đối tượng thử nghiệm một cách ngẫu nhiên
        cho treatment group (bị ảnh hưởng bởi hypothesised causal factor)
        or control group (không bị ảnh hưởng bởi hypothesised causal factor)
        
        A/B Testing
        
        Một đối tượng thử nghiệm được chỉ định ngẫu nhiên vào
        một trong các nhóm biến thể để so sánh xem nhóm biến thể nào hiệu quả hơn
        
        Observational Studies
        
        Đối tượng thử nghiệm chỉ được quan sát thụ động
        đo lường các biến quan tâm và xác định các mẫu hoặc mối quan hệ giữa chúng
        
        2 common methods
        
        Matching
        
        Multiple regression
      - 6 Criteria for Causation relationship
        
        Strength of association
        
        Consistency
        
        Specificity
        
        Temporality
        
        Gradient
        
        Plausibility
    - - Not cleaning the data prior to analysis
      - Not separating interaction effects
      - Not separating interaction effects
      - Relying on summary data
      - Creating data overload
  - - - Monitor data inputs as well as outputs and make adjustments as necessary
      - Refine the questions that we seek to answer
        through analysis over time
      - force us to confirm that do we answer right question or need to ask different questions
    - - Evaluate explanatory
        Act upon decision and evaluate outcome
        
        Ask explanatory
        Ask a specific question
        
        Collect explanatory
        Identify and collect relevant, quality data
        
        Analyse explanatory
        Conduct analysis of data to answer question
        
        Decide explanatory
        Interpret results and make decisions
        
        Communicate explanatory
        Communicate decisions
        
        1 more item...
      - Important attributes
        
        problem definition and framing
        
        decision-making methodologies
        
        evaluating and interating
    - - Observe
        
        Orient
        
        Decide
        
        Act
  - - - Ask questions?
        
        Does the data
        answer the original question?
        
        Yes
        
        How?
        
        No
        
        Is it because we don’t have the right data? or
        We haven’t got
        enough data?
        
        Does the data allow us to defend against any criticisms or objections?
        
        Are there any limitations on our conclusions, any angles we haven’t considered?
    - - Decision Type
        
        chia theo Frequency
        
        Non-programmed decision making
        
        Decisions that are not routine or may occured once
        
        Programmed decision making
        
        Decisions that are routine and scheduled
        
        chia theo Level
        
        Strategic
        
        Wide-affected to stakeholders decisions made by CEO,CFO,BOD
        
        Tactical
        
        Decisions affecting one or more subgroups
        made by middle/senior managers
        
        Operational
        
        Decisions with limited group
        day to day decisions
      - Decision making process
        
        Step 1: Evaluate situation
        
        Step 2: Analyse problem
        
        Step 3: Analyse decision
        
        Step 4: Analyse potential problems
- - - - How to create a Story with Data
        
        4 Steps
        
        Step 1.
        Define the goal
        
        Step 2.
        Setting the stage
        
        Step 3.
        Developing the story
        
        Step 4.
        Closing the story
        
        1 more item...
        
        make sure audience understand problem by using data to
        
        4 more items...
        
        More detail about the audience
        What is the level of domain knowledge?
        How much background information we need?
        
        Why are we telling this story?
        Who is the target audience?
        What do we want them to do?
    - - How to understand context
        
        Identify the audience
        
        Present the findings, not how to find them
        
        Plan the communication
    - - Các biểu đồ tương ứng với 4 ways to display data
        
        Only display
        raw material
        
        Single statistics (dữ liệu đơn)
        
        Tables (dữ liệu bảng)
        
        Charts (biểu đồ đơn)
        
        Comparing Magnitudes (Size)
        
        Bar charts
        (biểu đồ cột)
        
        Histograms
        (biểu đồ tần suất)
        
        Stacked bar charts
        (biểu đồ cột chồng)
        
        Grouped bar charts
        (biểu đồ nhiều cột)
        
        Part-to-whole charts
        (biểu đồ kết cấu)
        
        Pareto charts
        
        Funnel chart (biểu đồ phễu)
        
        Waterfall charts (biểu đồ thác nước)
        
        Area comparision chart
        (biểu đồ vùng)
        
        Proportional area chart
        
        Pictorial chart
        
        Tree map (Biểu đồ cây)
        
        Sankey diagram (biểu đồ tròn nhiều tầng)
        
        Desity and distribution charts (biểu đồ mật độ)
        
        Heat map
        
        Box and whisper plot
        
        One dimension scatter plot
        
        Geographical info charts (biểu đồ địa lý)
        
        Choropleth map
        
        Showing change
        over time
        
        Line chart (biểu đồ đường)
        
        Candlestick chart (biểu đồ nến)
        
        Slope graph (biểu đồ dốc)
        
        Layered area chart (biểu đồ lớp)
        
        Stacked area chart (biểu đồ lớp chồng)
        
        Stream chart
        
        Gantt chart
        
        Highlight Correlation
        
        Scatter plot (sơ đồ phân tán)
        
        Bubble chart (biểu đồ bong bóng)
    - - 2 useful approaches to support data visualisation
        
        Eliminate unnecessary details
        
        Data-to-ink ratio concept
        
        Data-to-ink ratio = (Data Ink) / (Data Ink + Redundant Ink)
        
        Data Ink: essential data
        
        Redundant Ink: can be removed from the visual,
        without reducing the effectiveness of the communication
        
        Ratio must close to 1.0 as much as possible
        
        Gestalt Principles
        
        6 nguyên tắc giải quyết tâm lý của nhận thức thị giác
        theo xu hướng increase power to eliminate unneccesary details
        
        Similarity
        
        1 more item...
        
        Draw attention to important info
        
        Using pre-attentive attributes
        
        Length& width
        
        Orientation
        
        Size
        
        Shape
        
        Enclosure
        
        Position
        
        Color Hue
        
        Color Intensity
        
        Add text decriptions and labels in diagrams/charts
    - - How to design a dashboard
        
        4 Steps
        
        Step 1: Choose thông số appropriate
        
        Select Visualisation
        
        There are more than 1 visualisation display
        Consider what fits space available
        
        Choose layout
        
        Group appropriate metrics and separate sections for clarify
        Reflect relative importance of different measures
        Review data to ink ratio
        
        1 more item...
        
        4 Final Consideration
        
        1 more item...
        
        Review business objectives?
        Consider if strategic, tactical or operational?
        Look at KPI for Input, Output, Efficiency, Quality, Outcome?
      - Types of dashboard
        
        Strategic Dashboards (For executives)
        
        Tactical Dashboards (For middle Management)
        
        Operational Dashboards (For Department manager)
      - Specified visualisation only for dashboard
        
        Bullet chart
        
        Gauges
        
        Progress bar
        
        Directional Indicators
  - - - finding patterns through breadth and depth of analysis
    - - effectively communicating the findings to decision makers
      - Benefits of data visualisation
        
        provide descriptive information
        
        provide the results of an analysis
        
        persuade others to act
        
        create an organisational memory