EPQ & HMM Research | Consumer Behavior Analysis

The Effect of Emotions on Consumer Purchasing Behaviour

Grade A* (Distinction) Stochastic Analysis Hybrid HMM-LR Model Predictive Modeling

This dissertation quantifies the "sensory marketing" phenomenon by decoding the temporal dynamics of consumer sentiment. By integrating Hidden Markov Models (HMM) to capture latent state transitions and Linear Regression for quantitative validation, the project provides a robust framework for predicting purchase intent within high-involvement market segments.

Technical Approach Key Findings

Latent State Transition Dynamics

HMM State Transition Logic

graph TD subgraph Hidden_States [Emotional Hidden States] S1((Happy/Joy)) S2((Surprise)) S3((Sadness)) S4((Anger)) end subgraph Transitions [State Transitions] S1 -- "P(ij)" --> S2 S2 -- "P(ij)" --> S1 S3 -- "Mitigated by Curiosity" --> S1 end subgraph Observations [Observable Outcomes] OB1[Purchase Decision] OB2[No Purchase / Browsing] end S1 -.->|High Prob| OB1 S2 -.->|Incentive| OB1 S3 -.->|Negative| OB2 S4 -.->|Inhibit| OB2 style S1 fill:#f9f1e7,stroke:#8B4513,stroke-width:2px,color:#8B4513 style S2 fill:#fdf8e4,stroke:#d4a373,stroke-width:2px,color:#8B4513 style S3 fill:#e9ecef,stroke:#adb5bd,stroke-width:1px,color:#495057 style S4 fill:#dee2e6,stroke:#6c757d,stroke-width:1px,color:#212529 style OB1 fill:#8B4513,stroke:#5d2e0a,color:#fff style OB2 fill:#6c757d,stroke:#495057,color:#fff style Hidden_States fill:#ffffff,stroke:#8B4513,stroke-dasharray: 5 5 style Transitions fill:none,stroke:none style Observations fill:none,stroke:none

*Stochastic mapping of Latent Emotional States to Observable Consumer Outcomes via Emission Probabilities.*

*Mapping non-observable emotional hidden states (categorized via Plutchik’s Wheel) to categorical purchase outcomes through a Markovian process.*

Hybrid Modeling Architecture

Quantitative Regression Analysis

By integrating a regression-based framework, the model assigns specific predictive weights to various emotional variables.

Impact Weighting: Determines exactly how much an increase in specific emotions, like Surprise, affects the final purchase probability.
Sample Size Validation: Utilizes power analysis for model validation to ensure statistical significance of regression coefficients.
Segmental Sensitivity: Evaluates how consumers with differing income levels respond to various marketing stimuli.

Hidden Markov Model (HMM)

The core engine utilizes an HMM to identify patterns in hidden emotional states that lead to specific consumer actions.

Transition Probabilities: Measures the likelihood of moving between emotional states, such as the stochastic shift from Expectancy to Joy.
Emission Matrix: Bridges latent emotional states to observable data points, including dwell time and navigational patterns.
EM Algorithm: References Expectation-Maximization to refine the parameters of the Markov chain choice model.

HMM Diagram Linear Model

HMM Model Visualization

Linear Regression Analysis

Empirical Research Insights

High-Conversion Catalysts

Transitions into Happy and Expectancy states demonstrate the strongest statistical correlation with immediate conversion decisions.

Negative State Mitigation

The research identified Curiosity as a critical mediator that can neutralize the deterrent effects of Anger on the purchasing path.

Temporal Trajectory

The sequence of emotions (e.g., Sadness → Surprise → Joy) is a more robust predictor of behavior than any single static measurement.

Impulse State Triggers

Unexpected rewards disrupt logical defense mechanisms, triggering high-frequency "impulse" states within the stochastic decision chain.

MENU