Please note that the contents of this offline web site may be out of date. To access the most recent documentation visit the online version .
Note that links that point to online resources are green in color and will open in a new window.
We would love it if you could give us feedback about this material by filling this form (You have to be online to fill it)



Frequently Asked Questions

  1. Getting Started
  2. Using the API
  3. Pricing
  4. Support and Feedback
  5. Storage and Content Policy

Getting Started

How do I sign up?
Please visit the Google Developers Console to sign up for prediction. Additionally, you must have an account with Google Cloud Storage to use the Google Prediction API.
What are some possible applications of the Prediction API?
The Prediction API can be used for many things. Application developers who want to predict new results from historic data can use our prediction capabilities. Example uses include user sentiment analysis, language identification, product recommendation, message routing, gene expression, and fraud detection, among others.
What training data does the Prediction API support?
Training data can be provided in one of three ways:
  • A CSV formatted training data file up to 2.5GB in size, loaded into Google Storage.
  • Training instances embedded directly into the trainingInstances parameter of the Insert request. Note: due to limits on the size of an HTTP request, this would only work with small datasets (< 2 MB).
  • Via Update calls - first an empty model is trained by passing in empty storageDataLocation and trainingInstances parameters into an Insert call. Then, the training instances are passed in using the Update call to update the empty model. Note: since not all classifiers can be updated, this may result in lower model accuracy than batch training the model on the entire dataset.
What functionality is available to developers and customers?
  • Discrete categories and continuously valued ranges as output,
  • Automatic selection from multiple machine learning methods,
  • Prediction from Google AppEngine and a variety of client libraries, and
  • A RESTful interface.
How accurate are the predictions?
The better your data, the better the predictions.

Using the API

Why can't I upload my files to Google Cloud Storage?
Make sure that you have an account with Google Cloud Storage: look on your Developers Console to see whether Google Cloud Storage is activated for your account. If not, sign up here .
How long does training take?
Training time varies with the size of the training data, it can go from less than a minute to up to a couple of hours.
How is estimated error calculated?
The API performs multiple rounds of cross-validation to calculate an estimated error for your dataset.
How long does prediction take?
Prediction usually takes less than 200ms.
Can I download my model?
You cannot currently download your model.
I get the message "ERROR" after training on my training file. What happened?
While we constantly strive to offer more detailed error messages, this message is currently being returned if training failed, the prediction engine can't find the training file, or for other errors. If you keep getting this message, check your training query to be sure that the path to your data is correct, and check your training data format to ensure that you do not have extra newline characters or commas, then try resubmitting your training request.
I'm getting a "404 File Not Found" error!
One possible reason for this is that you are using HTTP instead of HTTPS to call the Prediction API. The Prediction API supports only HTTPS access.
What happens if I exceed my prediction quota?
Please see our pricing page under "Paid Usage" for more details.
Can I use the Prediction API with other Google services?
Yes. For example, you can access the Prediction API from Google App Engine, or use the results from BigQuery as the basis for a predictive model. You are also required to use Google Cloud Storage in order to add training data to the Prediction API.
Does the API support unsupervised learning?
No. The Google Prediction API only supports supervised learning ; this means that the API can only learn by being told what the right answer is. The alternative, unsupervised learning , is not yet supported by Google Prediction.

Pricing

Where can I find pricing information?
Please read our pricing page for detailed pricing information.
What happens if I make fewer than 10,000 predictions in a month?
You will be charged our monthly access fee of $10, in addition to any charges incurred for any training you perform, after your six month trial period has expired.
What's an account?
An account is an individual project in the Developers Console using the Prediction API.
What happens after my free account has exceeded 20,000 queries or been active for more than six months?
To continue to use the API, your project will be required to sign up for a paid subscription to the Prediction API.
Are my Google Cloud Storage charges included?
Google Cloud Storage is billed independently of the Google Prediction API. Please visit the Google Cloud Storage pricing page for more details.
What about student/educator users?
Students and educators are welcome to use the free quota available to all new Prediction API users. In the future, we hope to better accommodate the needs of students and educators.

Support and Feedback

What are my support options?
We invite users to post their questions in our discussion forum . Our team will be actively monitoring your posts and answering questions, although we cannot guarantee a response. We understand that different customers have different needs and plan to announce additional support options in the future.
How do I request a feature?
Please see our suggestions, feedback, and notifications page .
How do I file a bug?
Please see our suggestions, feedback, and discussion page .

Storage and Content Policy

How and where is my data stored?
Your data is replicated in multiple data centers that are geographically distributed for greater availability. Your data is also replicated within each data center. Currently, data is stored in US data centers only. For security reasons, we do not disclose detailed location information about our data centers. We will inform users if we expand to locations outside of the United States.

Authentication required

You need to be signed in with Google+ to do that.

Signing you in...

Google Developers needs your permission to do that.