Discovering the Power of Text Topics and Path Analysis in SAS Visual Analytics
Introduction
SAS Visual Analytics offers a wide range of analytics objects that can be utilized for reporting, data analysis, and forecasting. These powerful tools help uncover patterns, trends, and insights within the data. In this post, we will be exploring how to use two key analytics objects—Text Topics (Sentiment Analysis) and Path Analysis—to enhance data-driven decision-making and extract valuable insights from your data.
|
Object Name |
Description |
Use Cases |
|
Text Topics |
A Text Topics object extracts and visualizes words from a document collection, identifies topics based on co-occurring terms, and analyzes sentiment as positive, negative, or neutral. |
Identify topics in customer reviews (e.g., "battery life," "customer service," "ease of use") to determine key areas for improvement. |
|
Path Analysis |
A path analysis object visualizes the flow of data between events, tracking the sequence and showing how data progresses through stages. This helps identify patterns, trends, and relationships between events. |
Helps visualize how data progresses through different stages in various processes. For example, in fraud detection, it shows how data flows from initial alerts through investigation and resolution stages. Similarly, in healthcare treatment, it tracks the movement of patient data from diagnosis to treatment pathways. This capability helps identify patterns and trends across different events, providing valuable insights into the flow of data and highlighting key relationships. |
Text Topics
Mission
- Identify topics in unstructured texts.
- Perform sentiment analysis on the texts.
- Create derived data item from the topics for further analysis.
Output Data Shape
A text topics object displays a set of words from a character data item that contains unstructured text. Here are the three main components:
- Bar Chart: Shows the topics identified from the documents and displays the document count for each topic.
- Word Cloud: Shows the relevant term for all documents. The size of each term in the word cloud indicates its importance.
- List Table: Shows the topic relevance and document details.
Key features
- Sentiment Analysis:
Just by selecting the Analyze document sentiment checkbox in the options pane for your text topic object. The system will perform the sentiment analysis automatically on your document provided.
Text Topics before the Sentiment Analysis
Text Topics after the Sentiment Analysis
In sentiment analysis, each word in the document is assigned a value based on the connotation of that word. The values are aggregated for the entire document to determine a sentiment score for the document. A sentiment score of less than 0.5 indicates a negative document, a score of 0.5 indicates a neutral document, and a score greater than 0.5 indicates a positive document.
When sentiment analysis is finished, the bar chart is updated to show the portion of negative, neutral, and positive documents in each topic.
- Derive Topics:
Just by right clicking the text topic object and selecting the Derive Topics, the VA can help user create new derived items in your data source. Two derived items for each topic can be created. One for topic itself and one for the relevance score.
Topic: The value for this derived item is either a 1 or a 0 to indicate whether each row contains the topic.
Relevance: The value for this derived item is between 1 and -1 to indicate the relevance score to the topic for each row.
After the configuration, user can see the newly created derived data items available in the Data pane.
Derived data items can be used to perform analysis on many different subjects. For example, user can assign derive topic as a group data role in a bar chart to see the distribution on topic for different category data item values.
Main Takeaways
- User can perform the text sentiment analysis simply by enabling the option.
- Derived data item is a powerful tool to be used for conducting further analysis on the text topics.
Path Analysis
Mission
- Visualize data flow: Path analysis helps you see how data moves through different stages, making it easier to track connections between events and values.
- Uncover patterns and trends: It reveals hidden relationships in the data, providing valuable insights for better decision-making.
- Optimize processes: Path analysis is useful for improving workflows, such as fraud detection or healthcare treatment, by identifying common paths and pinpointing areas for improvement.
Output Data Shape
The path analysis graph represents the customer journey through different stages of an eCommerce transaction.
- Stages of the Customer Journey:
- Browsing: Customers start by looking at products.
- Adding to Cart: They then add items to their cart.
- Checkout: Customers proceed to checkout.
- Payment Outcomes: At this point, the checkout process splits into three possible outcomes:
- Payment Failed: Transaction was unsuccessful.
- Payment Attempted: Customer tried to pay but did not complete the transaction.
- Payment Completed: Successful transaction.
- Representation of the Flow:
- The width of the blue paths represents the frequency of users moving from one step to another.
- The number at each stage (e.g., "5") indicates the count of users at that step.
- Drop Off:
- There is no drop-off indicated (0), meaning all users followed the defined paths.
Key features
- Data Roles:
- Event (Step): Defines each step in the process.
- Sequence Order (Timestamp): Ensures events occur in a logical sequence.
- Transaction Identifier (Session_ID): Tracks unique user sessions.
- Weight (Frequency): Determines how often each path occurs.
- Link Colors:
- Link color is based on the path they belong to, the events they originate from, or whether they are drop-off links.
- Path Display:
- The path display allows users to adjust the graph's appearance by selecting either the horizontal or vertical option.
Main Takeaways:
- All users successfully reached the checkout stage.
- Payment failures and incomplete attempts indicate possible issues in the payment process.
- The graph helps identify where users abandon their transactions and where improvements can be made in the checkout or payment process.
Conclusion:
In conclusion, text topic and path analysis in SAS Visual Analytics provide valuable insights into both event sequences and unstructured text data. Text topic analysis uncovers hidden trends in text sources, while path analysis helps track and visualize behaviors. These tools enable organizations to make data-driven decisions and optimize strategies based on a deeper understanding of their data.
Useful link:
Path analysis with SAS Visual Analytics
