Machine Learning Analysis
Machine learning is fast becoming a skill that is essential for efficient and effective analysis. Nanos has a tried and tested method for natural language processing and data mining using proprietary machine learning algorithms, yielding a more robust and better quality output with a faster turnaround.
Use in Quantitative Research
Nanos Research is now implementing machine learning algorithms to assist with the coding of open-ended questions. When dealing with an array of responses in large respondent sets, the algorithm allows us to group like responses automatically. Nanos begins this process by having 10% of the data set coded by an analyst and checked for accuracy. Nanos then applies a supervised model and codes the remaining responses with the coding key created by the analyst. This process enables Nanos to more quickly and accurately categorize sentiment and opinions, while minimizing human bias and error.
Nanos Research employs mixed-model algorithms to run machine learning ensembles. These are a mix of Hierarchical modelling, K-means clustering, and Latent Dirichlet Allocation modelling. This ensemble maximizes the efficiency and accuracy with which Nanos Research will be able to perform analysis, providing more robust results for qualitative research. This is accomplished by using algebraic formulas to measure the probability that words belong to certain topics, and how these topics relate to each other.
Use in Qualitative Research
In addition, Nanos is analyzing natural language using proprietary machine learning algorithms for qualitative research, including focus groups and elite interviews. This is ideal for the analysis of verbal responses, as the algorithm helps build a framework to provide replicable quantitative results when analysing qualitative data.
When dealing with an array of responses in a focus group or interview setting this utilizes the algorithm to group like responses, making it easier and more accurate for analysts to see themes and accurately categorize sentiment and opinions. Nanos uses this algorithm to analyse comments made during the focus groups and it adds another layer of depth and confidence to our analysis of the groups.
In this process, instead of the analyst having to make decisions about how to sort through the responses and discussion data, the algorithm decides the optimum way to organize the information available.