Business & Data Analytics Module 1

ITDX 41-Business & Data Analytics - Module 1 Introduction

• Business analytics (BA) encompasses skills, technologies, and practices for continuously developing insights into business performance, leveraging data and statistical methods.
• BA emphasizes exploring organizational data, focusing on statistical analysis, to support data-driven decision-making.

Significance of Business Analytics

• Data-driven decisions replace guesswork, improving decision-making.
• Converting data into valuable information.
• Gaining faster answers to business questions.
• Understanding customer behavior.
• Accessing key business metrics.

Uses of Business Analytics

• Improving business profitability.
• Growing market share and revenue.
• Enhancing shareholder return.
• Reducing overall costs.
• Maintaining a competitive position.
• Monitoring key performance indicators (KPIs).
• Responding to changing market trends in real-time.

Business Analytics Challenges

• Reliance on high-quality data volumes.
• Ensuring data quality.
• Increasing data storage needs for data warehousing.
• Influence of business analytics on customer interactions.
• The need for advanced technology infrastructure.
• Adapting to business changes brought by analytics.

BA Types

• Descriptive analysis.
• Predictive analysis.
• Prescriptive analysis.

Descriptive Analysis

• This analysis method examines past events to answer questions like "What has happened in the past?"
• Descriptive analysis summarizes and describes raw data for easier understanding.

Predictive Analysis

• Aims to forecast future events.
• Leverages forecasting techniques and statistical models.

Predictive Analysis Employ

• Relies on modeling and machine learning techniques.
• Machine Learning (ML): algorithms allowing computers to perform tasks without explicit instructions, using patterns and inferences.

Prescriptive Analysis

• Aims to provide advice on possible future outcomes, answering "What should we do?"
• Employs optimization and simulation algorithms.
• It allows users to recommend multiple solutions and guide decisions.

Few Software used for Business Analytics

• MS-Excel (spreadsheet)
• SPSS Statistics (statistical package)
• R (programming language)
• SAS (statistical analysis software)
• E-views (analytical tool)

Components of Business Analytics

• Data mining
• Text mining
• Forecasting
• Predictive analytics
• Optimization
• Visualization

Data Mining

• Discovers trends, patterns, and previously unknown relationships in vast datasets.
• Techniques like classification, regression, clustering, associations, and sequencing models are used.

Text Mining

• Extracts useful insights and patterns from textual data, extracting meaningful patterns and relationships from text collections.

Forecasting

• Analyzing and forecasting processes over time, using historical patterns to predict future outcomes.

Predictive Analytics

• Developing predictive scoring models (e.g., customer churn, credit scoring).
• Using simulations for optimal business decision-making.
• Providing visualizations of analysis results.

Optimization

• Use of simulations to identify best scenarios.
• Optimizing inventory levels.

Visualization

• Employing interactive and visual graphics for presenting insights from analysis results.

Data Mining as a Process

• Data integration and cleaning.
• Data pre-processing.
• Data selection and transformation.
• Data mining.
• Data evaluation and presentation.

Benefits of Data Mining

• Improved decision-making.
• Increased efficiency.
• Enhanced competitiveness.
• Improved customer service.

Disadvantages of Data Mining

• Excessive work intensity and staff training requirements.
• Large investments required for data collection and management.
• Security risks to data, including private customer details.
• Mistakes that can be made from inaccurate data.
• Managing large databases.

Big Data & Data Science

• Big data involves analyzing a huge volume of complex data in various formats.
• Data science uses advanced technologies like machine learning to analyze large datasets to gain insights for important decisions.

Source of Big Data

• Social media
• IoT devices
• Customer feedback
• E-commerce
• GPS data
• Machine-generated data
• Online shopping, emails etc.

Challenges with Big Data

• Data sharing and access.
• Privacy and security.
• Technical challenges like data quality, fault tolerance and scalability.
• Analytical challenges.

Data Science Process

• Business understanding
• Data understanding
• Data preparation
• Data exploration
• Model building
• Model evaluation
• Deployment

Steps for Data Science Processes

• Defining research goals
• Defining a clear research goal
• Understanding research context and mission
• Identifying analysis tools and resources
• Proof of concept or demonstrating work feasibility
• Identifying deliverables related to project objectives
• Defining project timeline and success metrics

Retrieving Data

• Finding data within a company can be challenging.
• Data may be stored in different repository forms (databases, data marts, warehouses, lakes).
• Company policies may require time to access data.

Cleansing, Integrating, and Transforming Data

• Data cleaning removes errors in data.
• Data is integrated to combine information from diverse sources.

Joining & Appending Tables

• Joining combines data between tables with shared values.
• Appending merges data from multiple sources together.

Handling Data Preparation steps

• Data is structured and processed for better model understanding before analysis.
• Variable elimination or reduction.
• Handling categorical and continuous data types is performed in the data preparation process.
• Numerical data type handling.
• Dummy variable creation to accommodate categorical variables for algorithms that cannot directly use categorical data types.

Exploratory Data Analysis (EDA)

• Analysis to identify trends, correlations, and outliers in the data.
• Use of graphical representations like plots and charts to aid data understanding.

Building Models

• Building and evaluating different models to find the best fit for prediction objectives identified earlier.
• Selecting predictive models.
• Testing and evaluation.

Presenting Findings

• Presenting models' results and findings to stakeholders, focusing on conveying complex data effectively.
• Decision-making and visualization.

Evaluation of Models

• Assessing if models successfully fulfil business success criteria.
• Review of the entire model-development process with summary of model findings and necessary corrections.
• Identifying subsequent steps based on model outcomes (deployment, further iteration, or new projects)

Data Mining - Architecture

• Front end
• Graphical user interface
• Pattern evaluation
• Knowledge Base
• Database server
• Data cleansing
• Data warehouse

Deployment

• Planning for deploying the models.
• Monitoring and maintaining operational phase of model.

Overfitting and Underfitting

• Overfitting: When a model learns too much from noise and inaccurate data. It performs well on training data but poorly on unseen data.
• Underfitting: when a model is too simple to capture the complexities of data. Perform poorly on training and unseen data.

Reasons for Underfitting

• Too simple model.
• Inadequate features,
• Insufficient data size.
• Excessive regularization
• Features not scaled.

Techniques to Reduce Underfitting

• Increasing model complexity
• Engineering more useful features.
• Handling Data Noise.
• Increasing training duration.

Reasons for Overfitting

• High variance, low bias.
• The size of the training data is too little.
• The model is too complex.

Techniques to Reduce Overfitting

• Improving data quality,
• Increasing data volume,
• Reducing model complexity,
• Early stopping in training,
• Regularization (Ridge and Lasso)
• Dropout in neural networks.