AWS Machine Learning RealTimePredictor returns UnknownoperationException in C#

Question

Using Visual Studio, and AWS .NET V 3.0.

I'm trying to perform a real-time Predict operation, and to verify the basic setup works, I first perform a GetMLModel() which works and returns the endpoint (Somewhere in the documentation is was mentioned to use that result as the service endpoint, but it's the same that is listed in the console). Is has status "READY", so far so good.

The exception occurs below on the line below "Prediction P = RTP.Predict(Data)". Data contains a Dictionary with all the prediction values.

Error: Error making request with Error Code UnknownOperationException and Http Status Code BadRequest. No further error information was returned by the service.

public static APIResult GetRealTimePrediction(Dictionary<string, string> Data, string PayloadJSON = null) {

        AmazonMachineLearningConfig MLConfig = new AmazonMachineLearningConfig();

        MLConfig.RegionEndpoint = Amazon.RegionEndpoint.USEast1;
        MLConfig.Validate();

        AmazonMachineLearningClient MLClient = new AmazonMachineLearningClient("xxx", "xxx", MLConfig);
        GetMLModelResponse MLMOdelResp =  MLClient.GetMLModel("xxx"); // <-- WORKS

        MLConfig.ServiceURL = MLMOdelResp.EndpointInfo.EndpointUrl;
        Console.WriteLine(MLConfig.ServiceURL);
        MLConfig.Validate();

        Amazon.MachineLearning.Util.RealtimePredictor RTP = new Amazon.MachineLearning.Util.RealtimePredictor(MLClient, "xxx");
        Prediction P = RTP.Predict(Data); // <----------------EXCEPTION HERE

}

(Obviously replace xxx with relevant values) :)

score 2 · Accepted Answer · answered Aug 31 '16 at 20:23

It turns out that this line:

MLConfig.ServiceURL = MLMOdelResp.EndpointInfo.EndpointUrl;

cases the MLConfig.RegionEndpoint to be reset. Even though the documentation indicates the RegionEndpoint can be determined from the ServiceURL (I'm pretty sure I read that), the RegionEndpoint needs to be set again before the RTP.Predict(Data) call.

Once I figured that out, I was able to reduce the code to just this, in case anyone else needs help. I guess adding too much information to the Configuration is NOT a good thing, as the AWS. NET library seems to figure all this out on its own.

public static APIResult GetRealTimePrediction(Dictionary<string, string> Data, string PayloadJSON = null) {
        AmazonMachineLearningConfig MLConfig = new AmazonMachineLearningConfig();
        MLConfig.RegionEndpoint = Amazon.RegionEndpoint.USEast1;
        MLConfig.Validate(); // Just in case, not really needed

        AmazonMachineLearningClient MLClient = new AmazonMachineLearningClient("xxx", "xxx", MLConfig);

        Amazon.MachineLearning.Util.RealtimePredictor RTP = new Amazon.MachineLearning.Util.RealtimePredictor(MLClient, "xxx");
        Prediction P = RTP.Predict(Data);
}

AWS Machine Learning RealTimePredictor returns UnknownoperationException in C#

1 Answers1