Predicting values - doubts regarding data collection

Hi there!

My name is Pedro Bertoleti and I’m a embedded software developer from Brazil. First of all, I’d like to thank you for providing such a good AI platform. Edge Impulse is truly awesome!

I wanna make a prototype of a selling/vending machine that involves predicting the sales revenue of the day, based on environmental variables (temperature and air humidity). I plan this prototype works like shown below:

  • I’ll collect data from its sales (date/time, what has been sold, minimum temperature, maximum temperature, average temperature, air humidity and revenue for this sell). These sales can happen any time, so there isn’t a “sample frequency” (as we usually have when sampling sensors data.
  • Once I have this complete dataset, I’ll split it into various datasets, accordingly to final price of sales. Then, I’ll have a dataset file for each final price.

Following this steps, am I able to do as described in if I consider sell revenues as labels and predict the final price of the day by informing temperatures (max, min and avg) and air humidity? Or isn’t this Edge Impulse approach adequate for what I wanna do?

In case applies to this, what should I put into CSV file timestamp column? I leave it as date/time?

Thanks in advance.

Best Regards,
Pedro Bertoleti

Hi @phfbertoleti With this I think you just want a CSV file like:


And label it 32.blah.csv (where 32 is your label). Then add a raw data block and a regression block, and train.

Hi @janjongboom. Thanks!

The question regarding timestamp was due to something that happened some time ago. As I remember, when i tried to upload a CSV file without timestamp column, this file was rejected by Edge Impulse. However, it was some time ago and I can’t cleary remember the errors I got at that time.

I’ll try as you said. Thanks again.

Yeah we updated this last week!


Hi @janjongboom. If you don’t mind, I’d like to ask another question.

I’ve observed that Edge Impulse always consider data over time (as expected in sensor analysis, for example, including the data for further regression). In my case, I wanna make temperature as independent variable and revenue as dependent variable. This way I can predict revenue as a function of temperature.

Is there any way to do this in Edge Impulse? Or can I only use time (or timestamp) as independent variable?

Thanks in advance.

Best Regards,
Pedro Bertoleti

@phfbertoleti it’s indeed not the thing that we built Edge Impulse for, but you can do it:

  1. Put the temperatures in CSV files, e.g.:


    And set to name to e.g. 230000.1.csv (where 230000 is the revenue).

  2. Upload the CSV files.

  3. Choose a ‘Raw Data’ processing block, and a regression learning block.

  4. Train.


Thanks for the help. It works!