This might be because the model you are evaluating is a regression model. Please note that BigML only provides the confusion matrix for classification models. Therefore you need to make sure the model you are evaluating is a classification model, and the objective field of your model is categorical.
You can easily convert your regression model into a classification model. Just discretize your objective field to convert the numeric fields into categorical fields. After that, make sure the raw undiscretized feature isn't included as an input to your model. Obviously, knowing the raw feature value will allow you to predict the discretized value with near-perfect accuracy, but this would be cheating as you won't have that information when making predictions with your model.
Please read chapter 10 of the Datasets documentation to learn how to add new fields to your dataset using different discretization strategies. Or click here to do it programmatically through the BigML API.