Downloaded Predictions do not Match Targets

[H. E. Downs]

I trained on the following data, and uploaded my Test File to see predictions from the model.  The Targets in my data are -1, 0, and 1.  But the predictions are fractional numbers, like 0.01, 0.50, etc.  I used all defaults on the first model, which had high confidence.  I am expecting the predictions to more or less match my original Targets.  What am I doing wrong?  

Training Data (80,000 rows, only showing Feature 31-35)
Training Data (showing last features and target)

Test Data (same, fewer rows)

Test Data for validation

Downloaded predictions from the model after uploading Test Data

Downloaded prediction file.

[H. E. Downs]

Modification:  I changed my Target to only predict 1 or 0 so it would run CLASSIFICATION models.  After training, I provided the same Test data and the prediction for all rows is 0.  This data is sparse, in that there may only be 10 or 20 occurrences of a specific  combination of features in the data that predicts a "1".  But there are many such combinations in the 80,000 rows.  What model should I use to find these sub-predictions?