Detailed Notes on Business Intelligence and Analytics Framework (BIAF)

1. Business Pressures - Responses - Support Model

 Business Environmental Factors: These are external factors that create pressures and
opportunities for organizations. Organizations respond to these factors to stay
competitive.
o Example:
 Pressure: A competitor launches a new product.
 Response: Your company invests in R&D to develop a better product.
 Support Model: Business Intelligence (BI) tools help analyze market
trends and customer preferences to guide R&D decisions.

2. Organizational Reprocessions
 Architecture of Business Intelligence (BI):
o Basic Components:
 Data Warehouse: A central repository where data from different
sources is stored. Think of it as a library where all books (data) are
organized and stored.
 Business Analytics: Tools and techniques used to analyze data. For
example, using analytics to predict future sales based on past data.
 Business Performance Management (BPM): A system that helps
organizations manage and measure their performance. For example,
tracking sales targets and comparing them with actual sales.
 User Interface: The front-end where users interact with the BI system.
It could be a dashboard that shows key performance indicators (KPIs)
like sales, revenue, and customer satisfaction.

3. High-Level Architecture of BI
 Diagram:

Data Source → Data Warehouse → Business Analytics → Performance & Strategy

o Data Source: Where data comes from (e.g., sales data, customer feedback).
o Data Warehouse: Data is cleaned, organized, and stored here.
 o Business Analytics: Data is analyzed to find patterns and insights.
o Performance & Strategy: Insights are used to make strategic decisions.

4. Components of BI Architecture
 BPM (Business Performance Management):
o User Interface: A dashboard that shows real-time data. For example, a sales
manager can see daily sales figures on their dashboard.
o Automated Decision Making: The system can make decisions based on
predefined rules. For example, if inventory levels are low, the system can
automatically reorder stock.

5. Types of BI
 Benefits of BI:
o Automated Decision Making (ADM): Reduces human error and speeds up
decision-making. For example, an e-commerce site automatically
recommends products based on a customer’s browsing history.
o Business Decision Rules (BDR): Customized rules for decision-making. For
example, a rule might state, "If sales drop by 10%, launch a promotional
campaign."

6. Analytical Processing & Transaction Processing

 OLAP Systems: Online Analytical Processing systems are used for complex queries.
For example, analyzing sales data across different regions and time periods.
 Technologies:
o Machine Learning (ML): Used for predictions. For example, predicting
customer churn based on past behavior.
o Big Data: Handling large datasets. For example, analyzing social media data to
understand customer sentiment.
o Forecasting: Predicting future trends. For example, forecasting next quarter’s
sales based on historical data.
o Data Constraints: Limitations in data processing. For example, missing data
can affect the accuracy of predictions.
7. Business Performance Management (BPM)
 Outcome of BI: BPM uses BI tools to add value to the organization. For example, by
identifying inefficiencies in the supply chain.
 Based on BI: BPM systems use data from BI to manage and measure business
performance. For example, tracking key metrics like revenue, profit, and customer
satisfaction.
 Tools:
o Enhanced Scorecard: A tool that tracks performance against strategic goals.
o Six Sigma: A methodology for process improvement.

8. BPM vs. BI
 BPM: An extension of BI that focuses on managing and improving business
performance.
 BPM = BI + Planning: Combines BI tools with strategic planning. For example, using
BI data to create a budget for the next fiscal year.

9. Performance Measurement
 Balanced Scorecard: A tool that helps managers track the implementation of
business strategy. It compares actual results against strategic goals.
o Example: A company’s strategic goal is to increase customer satisfaction. The
balanced scorecard tracks metrics like customer feedback scores and
complaint resolution times.
 Key Factors:
o Critical Success Factors: What to monitor. For example, sales growth and
customer retention.
o Strategic Goals and Targets: How to monitor. For example, setting a target of
10% sales growth and tracking progress monthly.

10. Variance Analysis

 Focus: Mostly on negative variances (failures to meet targets). For example, if sales
are 10% below target, the system flags it for review.
 Opportunities: Rarely reviews positive variances for potential opportunities. For
example, if sales are 15% above target, the system might suggest increasing
production to meet higher demand.
11. Data Types
 Cross-Sectional Data: A snapshot of data at a single point in time. For example, a
survey of customer satisfaction scores taken in January.
 Time Series Data: Data collected over time. For example, monthly sales data for the
past 5 years.
o Seasonality: Short-term cyclic changes. For example, ice cream sales peak in
summer.
o Cyclic: Long-term cyclic changes. For example, economic recessions every 10
years.
o Irregular: Random events. For example, a sudden spike in sales due to a viral
marketing campaign.

12. Time Series Forecasting

 Objective: Forecast future values by decomposing data into trend, seasonality, and
irregular components.
 Methods:
o Exponential Smoothing: A technique for smoothing time series data. For
example, predicting next month’s sales based on a weighted average of past
sales.
o ARIMA Models: AutoRegressive Integrated Moving Average models are used
for forecasting. For example, predicting stock prices based on past trends.

13. ARIMA Models

 Steps:
1. Model Identification: Use ACF (AutoCorrelation Function) and PACF (Partial
AutoCorrelation Function) to identify the model.
2. Parameter Estimation: Fit the ARIMA model (AR(p), I(d), MA(q)).
3. Model Diagnostics: Check the accuracy of the model.
4. Forecasting: Generate forecasts and confidence intervals.

14. Key Concepts in ARIMA

 AR(1) Model: yt=ϕyt−1+ϵtyt=ϕyt−1+ϵt
 o Example: If this month’s sales are highly dependent on last month’s sales, we
use an AR(1) model.
 MA(1) Model: yt=θϵt−1+ϵtyt=θϵt−1+ϵt
o Example: If this month’s sales are affected by last month’s random shock
(e.g., a sudden market crash), we use an MA(1) model.
 ARMA(1,1) Model: Combines AR and MA components.
o Example: If sales are affected by both last month’s sales and last month’s
random shock, we use an ARMA(1,1) model.

15. Model Diagnostics

 Coefficient Significance: If the coefficient is greater than 2x the standard error, it is
significant.
o Example: If the coefficient is 0.5 and the standard error is 0.1, then 2x the
standard error is 0.2. Since 0.5 > 0.2, the coefficient is significant.
 Log Transformation: Used to stabilize variance in time series data.
o Example: If sales data has increasing variance over time, taking the log of
sales can stabilize the variance.

16. Summary
 BI and BPM: Tools for managing and measuring business performance.
 Time Series Analysis: Essential for forecasting and understanding trends.
 ARIMA Models: Powerful tools for time series forecasting, combining autoregressive
and moving average components.

Diagrams
1. High-Level Architecture of BI

Data Source → Data Warehouse → Business Analytics → Performance & Strategy

2. ARIMA Model Process

1. Model Identification → 2. Parameter Estimation → 3. Model Diagnostics → 4. Forecasting

3. Time Series Decomposition
Time Series = Trend + Seasonality + Irregular

These detailed notes, along with examples and diagrams, should make the concepts of
Business Intelligence, Business Performance Management, and Time Series Analysis easier
to understand and apply.

Auto Correlation Function (ACF)

 Autocorrelation measures the correlation between a time series and its lagged
versions. For example, how today’s stock price is related to yesterday’s price, or the
price from two days ago.
 ACF plots the correlation coefficients for different lags (time intervals).
 Measures the correlation between the time series and its lagged versions.
 Helps identify MA (Moving Average) components.
 If ACF cuts off at lag q, the process is likely MA(q).

Partial Autocorrelation Function (PACF)

 PACF measures the correlation between a time series and its lagged versions,
but after removing the influence of intermediate lags.
 For example, PACF at lag 2 measures the correlation between the time series and its
value at lag 2, excluding the effect of lag 1.
 Shows the direct correlation between a time series and its lags, removing effects of
intermediate lags.
 Helps identify AR (AutoRegressive) components.
 If PACF cuts off at lag p, the process is likely AR(p).

Autoregressive (AR) Model

 An AR model predicts the future value of a time series based on its past values.
 The model is defined as:
Xt=c+ϕ1Xt−1+ϕ2Xt−2+...+ϕpXt−p+ϵtXt=c+ϕ1Xt−1+ϕ2Xt−2+...+ϕpXt−p+ϵt
o XtXt: Current value
o Xt−1,Xt−2,...Xt−1,Xt−2,...: Past values
o ϕ1,ϕ2,...ϕ1,ϕ2,...: Coefficients
o ϵtϵt: Error term
  It uses the number of past values to predict the current value. For example, AR(1)
uses only the previous value, AR(2) uses the previous two values, etc.

Moving Average (MA) Model

 An MA model predicts the future value of a time series based on past error
terms (residuals).
 The model is defined as:
Xt=μ+ϵt+θ1ϵt−1+θ2ϵt−2+...+θqϵt−qXt=μ+ϵt+θ1ϵt−1+θ2ϵt−2+...+θqϵt−q
o μμ: Mean of the series
o ϵtϵt: Current error term
o ϵt−1,ϵt−2,...ϵt−1,ϵt−2,...: Past error terms
o θ1,θ2,...θ1,θ2,...: Coefficients
 It uses the number of past error terms to predict the current value. For example,
MA(1) uses only the previous error term, MA(2) uses the previous two error terms,
etc.

Auto Regressive Moving Average Model (ARMA)

 ARMA combines both AR and MA models. It uses past values and past error terms to
predict the future value.
 The model is defined as:
Xt=c+ϕ1Xt−1+ϕ2Xt−2+...+ϕpXt−p+ϵt+θ1ϵt−1+θ2ϵt−2+...+θqϵt−qXt=c+ϕ1Xt−1+ϕ2Xt−2+...
+ϕpXt−p+ϵt+θ1ϵt−1+θ2ϵt−2+...+θqϵt−q
 ARMA is useful for stationary time series (no trend or seasonality).
 p: Order of the AR component
 q: Order of the MA component

Auto Regressive Integrated Moving Average Model

(ARIMA)
 ARIMA is an extension of ARMA that can handle non-stationary time series (data
with trends or seasonality).
 It includes an integration (I) step to make the data stationary by differencing
(subtracting past values from current values).
  ΔdXt=c+ϕ1ΔdXt−1+ϕ2ΔdXt−2+...+ϕpΔdXt−p+ϵt+θ1ϵt−1+θ2ϵt−2+...+θqϵt−q

 ΔdXt: Differenced time series (d is the order of differencing)

 p: Order of the AR component
 d: Order of differencing
 q: Order of the MA component

Summary of Key Concepts:

Mode Parameter
Description Use Case
l s

AR Uses past values

pp (AR Stationary data
order)

MA Uses past errors

qq (MA Stationary data
order)
Combines AR and
ARMA
MA
p,qp,q Stationary data
ARIM ARMA + Non-stationary
A differencing
p,d,qp,d,q data