2 Regression Analysis models relationships between variables, examining how independent variables influence a dependent variable.

e. This is vital for businesses to

predict outcomes like sales based on factors like marketing spend. Linear regression, establishing a straight-line relationship, is common due to its simplicity,
though non-linear models exist for complex patterns. Logistic Regression is specialized for binary dependent variables (e.g., ‘yes’/’no,’ ‘churned’/’retained’),
predicting one of two outcomes.
3 In Economics, regression quantifies price-demand relationships and aids macroeconomic forecasting (e.g., GDP). Finance uses it for revenue forecasting (linking
ad spend to sales) and evaluating bank performance by isolating variables’ impact on metrics. It also helps understand customer loyalty drivers by linking
satisfaction to retention. Accounting employs it for cost analysis, separating fixed and variable costs against activity levels (like units produced) to predict future
costs and identify efficiencies.
4 Classification Tree Analysis uses a tree-like model to map decisions and assign observations to predefined categories. It works by recursively partitioning data into
increasingly homogeneous subsets based on decision rules applied to independent variables. Its strength lies in its intuitive, visual flowchart structure, making it
highly interpretable. Predictions involve traversing the tree from the root to a terminal (leaf) node, which assigns a class label. Overfitting, where the model
learns training data noise, is a risk, mitigated by pruning or using trees in ensemble methods like Random Forests.
5 In Finance, classification trees aid customer segmentation (e.g., ‘high-value investor,’ ‘potential churner’) for targeted marketing. Economists use them to classify
firms by default risk or regions by economic development stage. In Accounting, they help auditors identify high-risk transactions by learning from past
problematic entries, focusing audit efforts. They can also streamline internal processes like expense approvals by automatically classifying submissions based on
policy and historical data.
6 Machine Learning (ML), a field in AI, enables systems to learn from data without explicit programming, identifying patterns to make predictions or automate
decisions. Computers autonomously improve performance by analyzing vast data. Common types include supervised learning (from labeled data), unsupervised
learning (finding patterns in unlabeled data), and reinforcement learning (learning via rewards/penalties). ML models are adaptable and scalable but some
advanced models (e.g., deep learning) can be “black boxes,” with opaque decision-making processes.
7 In Finance, ML revolutionizes fraud detection, powers algorithmic trading, and enhances credit risk assessment using diverse data. Economists leverage ML for
sophisticated forecasting, modeling complex, non-linear relationships in large datasets (including unstructured data like news sentiment) for predictions like
inflation or market movements. Accountants benefit from ML automating voluminous data scanning in audits, identifying anomalies indicative of errors or fraud,
improving audit efficiency and effectiveness.
8 Genetic Algorithms (GAs) are optimization methods inspired by natural selection, finding near-optimal solutions for complex problems where traditional methods
are impractical. They use a probabilistic search guided by three main operators: selection (choosing “fitter” parent solutions), crossover (combining parent
solutions to create offspring, inheriting traits), and mutation (introducing small, random changes to maintain diversity and explore the solution space), iteratively
refining a population of potential solutions.
9 In Finance, GAs optimize algorithmic trading strategies by evolving trading rules and refine fraud detection models by finding optimal combinations of predictive
variables. Accounting uses GAs for complex distribution, resource allocation, and scheduling challenges to minimize costs or time, and to enhance bankruptcy
prediction models by selecting informative financial ratios. Economists apply GAs to model adaptive economic behavior (e.g., cobweb model) and optimize
economic policy parameters or calibrate agent-based models.
10 Sentiment analysis enables organizations to monitor public discourse about their brand, products, and services, often in real-time. This is crucial for
understanding customer sentiment and identifying early signs of dissatisfaction. Using AI, these tools distill customer opinions from unstructured data (social
media, reviews) minimizing human bias. This real-time monitoring, especially on social media, empowers proactive crisis management and reputation
safeguarding. Different types of sentiment analysis cater to specific analytical needs.
11 Key approaches interpret emotional tone in text. Rule-Based Analysis uses predefined lexicons (dictionaries of words with sentiment scores) and grammatical
rules to calculate overall sentiment (e.g., “affordable” is positive, “slow” is negative). Machine Learning (ML) Analysis trains algorithms on large, labeled datasets
(text with known sentiments), allowing the system to learn patterns and contextual cues to predict sentiment in new text. The Hybrid Approach combines rule-
based techniques and ML models, aiming for improved accuracy by leveraging both explicit knowledge and pattern recognition, though it can be complex to
develop.
12 Sentiment analysis faces challenges as an evolving technology. Lack of contextual understanding is significant; tools struggle without broader situational
awareness, often requiring pre-processing or manual correction. Detecting irony and sarcasm is difficult, as these rely on non-textual cues, leading to
misinterpretations (e.g., “Awesome, another parking ticket” read as positive). Complex negation (e.g., “I wouldn’t say it was cheap”) can be misread if nuances
are missed. Finally, idiomatic language (“beat around the bush”), where literal meaning differs from intended meaning, poses a considerable challenge, affecting
accuracy.
13 Sentiment analysis provides actionable business intelligence. In Market Analysis, it gauges market mood and anticipates shifts by examining sentiment across
social media and news, uncovering trends. For Customer Analysis, it offers insights into perceptions of a business and its offerings, informing strategy, resource
allocation, and product adjustments, enabling early detection of concerns. In Mergers and Acquisitions (M&A), it provides insights into public, investor, and
stakeholder perceptions of deals, and can monitor competitor reactions, offering a holistic view of a transaction’s potential impact.
14 CommBank, serving over 10 million customers, faces challenges analyzing vast customer feedback to address dissatisfaction, improve services, and boost
retention. Implementing Sentiment Analysis allows automated processing of unstructured feedback from diverse channels (social media, reviews, surveys, call
transcripts, emails), classifying it as positive, negative, or neutral.
This helps CommBank quickly spot dissatisfaction trends (e.g., mortgage process issues, fee concerns, support wait times) for targeted responses. Analyzing
sentiment around campaigns or product launches allows real-time strategy adjustments. Benchmarking public perception against competitors identifies market
leadership areas and improvement opportunities. Ultimately, this enhances customer experience, informs decisions, and protects brand reputation, driving
loyalty.
15 For CommBank, Sentiment Analysis limitations include:
Context/Nuance: Basic tools might identify negative sentiment but not the specific reason (e.g., mortgage times vs. online banking confusion), hindering precise
action.
Sarcasm/Cultural Nuances: Sarcastic praise (“fantastic service” for a long wait) or culturally specific expressions from CommBank’s diverse customers can be
misclassified.
Domain-Specific Language: Financial jargon (“LVR,” “offset account”) may be misunderstood by generic tools lacking specialized financial lexicons.
 Signal vs. Noise: Distinguishing actionable feedback from irrelevant mentions in high-volume data is vital; a negative spike could be misleading.
Actionable Insights: Overall scores (e.g., “70% positive”) don’t inherently dictate actions; human interpretation is crucial for strategies.
16 CommBank’s use of Sentiment Analysis raises:
Ethical Issues:
Transparency & Trust: Customers should be informed if feedback is analyzed; handling opinions carefully maintains trust.
Algorithmic Bias: Models might reflect training data biases, misinterpreting sentiment from certain groups, leading to inequitable service.
Accuracy: Misinterpretations (context, sarcasm, jargon) can create flawed data, leading to poor decisions.
Legal & Privacy Issues:
Australian Privacy Principles (APPs): Data collection/use must adhere to the Privacy Act 1988 regarding personal information handling.
Data Security & PII Anonymization: CommBank must protect Personally Identifiable Information (PII) with robust security and anonymization.
Purpose Limitation & Consent: Data collected for one purpose shouldn’t be broadly used for sentiment analysis without appropriate disclosure/consent.