Survey analysis process for extracting and organizing dynamic textual content to use as input to structural equation modeling (SEM) for survey analysis in order to understand how customer experiences drive customer decisions

A survey analysis process for extracting and organizing dynamic textual content to use as input to structural equation modeling (SEM) for survey analysis in order to understand how customer experiences driver customer decisions, said process comprising:

• cleaning raw input data textual words;
• mapping the clean textual words into numeric vector representations;
• clustering words based on the numeric vector representations into a number of word clusters using a Gaussian mixture model;
• converting the input data textual words into a numeric vector by counting the occurrence of words in each word cluster;
• clustering open-ended questions into a number of sentence clusters using K Means clustering;
• defining the topics of each sentence cluster by the most frequent words;
• creating master data frame where all topics extracted from previous exercise as columns and comments as rows and setting topic column to one of a zero binary value when the sentence includes no words associated with the topic and a one binary value when the sentence includes a word associated with the topic;
• developing a first iteration of a theoretical SEM diagram where a group of topics that belong to a similar category are programmatically loaded for the category;
• using recorded questions as the dependent variables for the structural equation model;
• running the structural equation model and finding statistically significant coefficients for each topic category independent variable on recorded question dependent variables;
• further refining the structural equation model to eliminate topic categories;
• directly loading the topics on each question to identify major topics for each question separately;
• and
• reporting topics with statistically significant impact on each question to elaborate how the trends in those questions are related to the defined topics.