In the rapidly evolving landscape of the electric car industry, understanding user demands is crucial for driving product innovation and market success. Traditional methods often fall short in capturing the dynamic and multifaceted nature of these demands, especially in fast-paced environments like social media. In this study, we leverage the BERTopic model to analyze user discussions from a popular social platform focused on the electric car segment, specifically examining topics related to a prominent model. By integrating Maslow’s Hierarchy of Needs, we explore the hierarchical relationships among user demands and track their evolution over time using dynamic topic modeling. This approach allows us to uncover latent themes, assess their relevance, and identify trends that can inform strategic decisions for electric car manufacturers.
Our methodology begins with data collection and preprocessing, where we gather user-generated content related to electric car discussions. We employ advanced natural language processing techniques to clean and tokenize the text, ensuring high-quality input for topic modeling. The BERTopic model is then applied, utilizing pre-trained language models to generate semantic embeddings, reduce dimensionality, and perform clustering. This process enables us to identify key topics that represent user interests in the electric car domain. For instance, we extract themes such as new model releases, design aesthetics, and performance metrics, which are central to electric car adoption.
To quantify the importance of terms within each topic, we use the class-based TF-IDF (c-TF-IDF) method, which refines traditional TF-IDF by incorporating class-specific information. The formula for c-TF-IDF is given by:
$$W_{x,c} = \text{tf}_{x,c} \times \log \left(1 + \frac{A}{\|f_x\|}\right)$$
where \(W_{x,c}\) represents the weight of term \(x\) in cluster \(c\), \(\text{tf}_{x,c}\) is the frequency of term \(x\) in cluster \(c\), \(A\) is the average number of terms per cluster, and \(\|f_x\|\) denotes the frequency of term \(x\) across all clusters. This approach enhances the representation of topics in electric car discussions by emphasizing terms that are distinctive to each cluster.
In our analysis, we identify several hotspot topics that dominate user conversations about electric cars. The table below summarizes the top topics based on document frequency, highlighting the core themes and their characteristic terms:
| Topic ID | Topic Name | Characteristic Terms | Document Frequency (%) |
|---|---|---|---|
| 0 | New Model Launch | electric car, model, release, battery | 18.5 |
| 1 | Design Aesthetics | design, style, interior, color | 15.2 |
| 2 | Sales Performance | sales, growth, market, demand | 12.8 |
| 3 | Brand Comparisons | comparison, brand, competitor, features | 10.4 |
| 4 | Technology Integration | tech, system, update, connectivity | 9.7 |
These topics reflect the diverse interests of electric car users, ranging from practical aspects like sales and technology to emotional elements like design. For example, the “Design Aesthetics” topic often includes discussions on cultural elements, such as traditional patterns, which resonate with users’ identity and values. This highlights how electric car manufacturers can leverage aesthetic appeal to enhance user engagement.
Next, we examine the correlations between topics using cosine similarity, which measures the directional similarity between topic vectors. The cosine similarity between two vectors \(\mathbf{A}\) and \(\mathbf{B}\) is defined as:
$$\cos(\theta) = \frac{\mathbf{A} \cdot \mathbf{B}}{\|\mathbf{A}\| \|\mathbf{B}\|}$$
where \(\mathbf{A} \cdot \mathbf{B}\) is the dot product of the vectors, and \(\|\mathbf{A}\|\) and \(\|\mathbf{B}\|\) are their magnitudes. This metric helps us identify clusters of related topics, which we then map to Maslow’s Hierarchy of Needs to understand the underlying psychological drivers. The hierarchy includes physiological, safety, social, esteem, and self-actualization needs, and we find that electric car user demands often align with these levels. For instance, safety needs correspond to topics like battery reliability and crash tests, while self-actualization relates to advanced features like over-the-air updates and autonomous driving in electric cars.
The following table illustrates the correlation analysis between selected topics, showing how closely they are related and their associated Maslow levels:
| Topic Pair | Cosine Similarity | Maslow Level | Interpretation |
|---|---|---|---|
| Technology Integration & Safety Features | 0.88 | Safety Needs | Users associate tech updates with enhanced safety in electric cars. |
| Design Aesthetics & Social Recognition | 0.85 | Esteem Needs | Design elements fulfill desires for status and identity in electric car communities. |
| New Model Launch & Self-Actualization | 0.82 | Self-Actualization | Latest electric car models represent personal achievement and innovation. |
This analysis reveals that electric car demands are hierarchical, with lower-level needs like safety serving as foundations for higher-level aspirations like self-expression. For example, users who discuss battery safety (a safety need) are more likely to engage in topics about customizing their electric car (an esteem need), indicating a progression in demand fulfillment.
To track the evolution of these topics over time, we apply the Dynamic Topic Model (DTM) component of BERTopic, which segments the data into quarterly intervals. This allows us to observe how user interests in electric cars shift in response to external factors such as industry events, policy changes, and technological advancements. The evolution trends are visualized through topic frequency plots, showing peaks and troughs that correlate with real-world occurrences. For instance, a surge in discussions about electric car autonomy often follows major tech announcements, while sales topics spike during promotional periods.
We also compute the temporal changes in topic prominence using a time-series formula. The frequency of a topic \(T\) at time \(t\) can be modeled as:
$$F_T(t) = \alpha \cdot \sum_{i=1}^{n} w_i \cdot e^{-\beta (t – t_i)^2}$$
where \(F_T(t)\) is the frequency of topic \(T\) at time \(t\), \(\alpha\) is a scaling factor, \(w_i\) represents the weight of event \(i\), \(t_i\) is the time of event \(i\), and \(\beta\) controls the decay rate of event influence. This equation helps quantify how specific events, like the launch of a new electric car model, affect user attention over time.
Our findings indicate that electric car user demands are highly dynamic, influenced by factors such as:
- Industry Innovations: Breakthroughs in battery technology or autonomous features lead to sustained interest in related topics.
- Market Competition: As more players enter the electric car market, users become more rational, focusing on性价比 (value for money) and comparative analyses.
- Brand Strategies: Marketing campaigns and partnerships (e.g., with tech companies) shape user perceptions and discussions, often elevating topics related to brand identity and loyalty.
In conclusion, the BERTopic model proves effective in capturing the complex and evolving nature of electric car user demands. By combining topic modeling with psychological theory and dynamic analysis, we provide a comprehensive framework for understanding what drives consumer behavior in this sector. This approach not only identifies current hotspots but also predicts future trends, enabling electric car manufacturers to align their strategies with user expectations. For instance, emphasizing safety and design can address foundational needs, while highlighting innovation can cater to higher-level aspirations, ultimately fostering greater adoption and satisfaction in the electric car market.
Future work could expand this analysis to include multimodal data from various platforms, further refining the understanding of electric car demands across different contexts and cultures.