In this article, we'll show you how to start using generative AI for data exploration and analysis, from creating a data set in the Attune platform, to importing it into a generative AI tool and creating prompt to manipulate the data and create analysis.
In this example, we will be using Chatgpt from Open AI as it is currently the most advanced tool to generate and run analysis. However, by the time you read this there may be other tools available. In any case, the logic will remain the same.
Generative AI tools will be analyzing the data by creating and running Python code on your behalf. Python is a programing language - but don't worry, you don't need to know anything about it. All you need to do is make sure whatever results it outputs make sense because sometimes it is wrong, which is a phenomenon called 'AI hallucination'.
An important thing to keep in mind is that if you are using free accounts, that data you are entering may be used for future training so we highly recommend selecting a plan where you can disable chat training. Please contact us if you need help figuring out the right plan for you.
Here are a few steps with prompts you can enter. The data file we used for this example is downloadable from the link at the end of this help article.
1. The first step is to decide what dataset you will want to analyze. Refer to the article How to generate a CSV for AI Data Exploration for more info on how to create a file for AI analysis.
2. Upload the file to a Generative AI tool and enter the following prompt:
Here is a spreadsheet containing time in the first column, building name in the second column, sensor location in the third column, and then the rest of the columns are sensor values for each set of building name and sensor location for different timestamps. Analyze the spreadsheet and let me know any issues before we move on to further data analysis.
3. Look at the response and see if there are any specific steps that need to be taken to clean the data, such as missing data, etc. There is nothing to do in our case. What you see may differ from the screenshot below as each prompt will result in different responses (the numbers will remain the same).
4. If you want to take a quick look at the data, you can enter the following prompt:
For each type of data, draw a line plot showing the data from the different locations
5. To filter the data to only values during operating hours, you can enter the following prompt. If you are analyzing data from buildings in different timezones, you can clarify which building is in which timezone.
We want to focus the analysis inside operating hours. Filter the data to only keep data Monday-Friday from 9am to 5pm ET. All the data is already in ET timezone
6. A simple analysis that can then be done is to look at the percentage of time values are below a defined threshold. This can be used to find rooms with issues or rank different rooms/locations to prioritize retrofit work.
For each sensor in each location, calculate the percentage of time below threshold during the operating hours. For CO2, use 1000ppm as the threshold; for Temperature, use 76F; for humidity, use 75%; and for TVOC, use 2200ppb
7. Then:
Looking at CO2 only, order the locations starting with the one with the least percentage of time below threshold
This is a simple example that shows how simple it is to answer questions that would take a long time to work on using excel. More complicated analysis are possible, just keep in mind:
Break down the questions/tasks in as many steps as possible as generative AI is not good at executing a lot of tasks at once
Between each step, verify that the results make sense and feel free to ask clarifying questions or highlight any inconsistencies
Generative AI is evolving very rapidly and results may be inconsistent between working sessions
If there is too much time between prompts or if the session is very long, there are currently issues where the tool won't be able to continue generating analysis. A good practice is to ask for downloadable spreadsheets after different steps so you can restart from there at any point.
Feel free to contact us with any questions!