Ayasdi Core is a unique analytics tool providing topological data analysis
Rating: 7 out of 10
IncentivizedUse Cases and Deployment Scope
We used Ayasdi Core to analyze the correlation between surgical procedures and patient outcomes in our surgery department. We specifically wanted to find major factors (physician demographics, unit, facility, treatment, etc.) affecting the length of stay of patients. Our goal was to find and control the factors so that we can minimize length of stays without negative effects to other clinical outcomes. We extracted data sets from our clinical databases and fed them to Ayasdi. A team was formed with members of domain experts including physicians, nurses, administrators, and informaticists. We conducted the analysis using Ayasdi Topological Data Analysis and visualizations, the result was presented internally in our organization.
Pros
- Ayasdi Core provides an easy way to get some insight on data. Typically analytics may require having a model or hypothesis before starting to look into the data, but Ayasdi lets you just feed the data first then start seeing what the data looks like.
- Ayasdi Core's topological network visualization is quite unique. It allows you to explore patterns and potential relations between multiple data elements. A user can also dynamically navigate data with different aspects on the web.
- The Web version of Ayasdi is easy to use, stable, and fast. It hasn't crashed even when we feed it a lot of data sets, although it took time.
Cons
- Use of Python SDK is required to feed data into Ayasdi, but it lacks training materials or sample codes for a novice to get started.
- Although Web UI of Ayasdi is looking good, often it freezes when the user runs an analysis. It doesn't crash but the web page needs to be refreshed to see the progress of analysis.
- Algorithms provided by Ayasdi, such as metrics types, lens types need to be explained (what they are and what their strengths and weaknesses are). We had to Google or do research on our own to understand what they are.
Likelihood to Recommend
Well suited:
1. If data set is not yet well organized.
2. Hypothesis is not yet established.
3. Need to visually explore to find patterns of data (often when analysts have no good understanding of data)
4. When [you need] to analyze events with a timeframe (specifically a sequence of events as a transaction)
Less appropriate
1. If a data set is very large, such as Hadoop data, it becomes hard to manage data pipeline and process to feed the data into Ayasdi. To be feed into Ayasdi, data should be aggregated or organized to some level.
1. If data set is not yet well organized.
2. Hypothesis is not yet established.
3. Need to visually explore to find patterns of data (often when analysts have no good understanding of data)
4. When [you need] to analyze events with a timeframe (specifically a sequence of events as a transaction)
Less appropriate
1. If a data set is very large, such as Hadoop data, it becomes hard to manage data pipeline and process to feed the data into Ayasdi. To be feed into Ayasdi, data should be aggregated or organized to some level.