Reddemy may receive an affiliate commission if you enroll in a paid course after using these buttons to visit Udemy. Thank you for using these buttons to support Reddemy.
Reddit Posts and Comments
0 posts • 17 mentions • top 3 shown below
2 points • inspiredDeveloper
1 points • GantMan
This course jumps into the ground up approach. Honestly my course is a bit more hands on.
1 points • wildcamp
> How would I go about doing this?
Basically it's a classification problem. You are classifying responses into valid or invalid.
First you'll need some training data - which means the form needs to exist and have been used and you have some data collected before you can add the machine learning element.
Once you have some sample responses (say 100) you can manually tag them with an valid / invalid label. Then you need to identify some features of the data you can use as inputs for the machine learning model - for example, time spent on form, or length of a text field. Once you have this, you need to define your model structure and can train it.
The details are more complex but there are many resources to learn such as this TensorFlow.js course which covers the basic concepts.
> Also, how would I go about training the model? Would I train it once or continue to train it progressively somehow?
Training progressively will make results more accurate, but of course it's more complex as it would involve labeling more data as valid/invalid.
> Would creating an endpoint in AWS be better?
You could use any provider that allows you to use one of the TensorFlow.js Node modules. An Express app should work fine, serverless function solutions might also work fine but the best way to find out is to try it!
> Will running it in tensorflowjs cause heroku to overcharge me? Seeing as it’s quite computive intensive.
Not sure about that - but if concerned about usage costs you could instead rent a VPS or dedicated server so you'd know for sure what you're going to pay.