In Azure Machine Learning Studio, I tried one experiment to predict fertility diagnosis as normal or abnormal as an output. I have a dataset in .csv file of people who are normal or altered. Following isa screenshot of that dataset having information like seasons in which analysis was performed, age at the time of analysis, childhood diseases which affect fertility process, accidents, serious trauma history if any, high fevers in the last few years, frequency of alcohol consumption, smoking habit, number of hours spent sitting per day.
Fertility_diagnosis.csv
![]()
I created machine learning experiment in Azure ML studio, I used splitter to split data into 70% - 30%; 70% of the data I passed for model training purposes and 30%of the data I kept for score model, I passed Two class Boosted Decision classification algorithms to train the model. To choose an algorithm you require algorithm knowledge; if you are beginner in Machine learning experiments this is advice You should pass one by one classification algorithms in training experiments and you should check ROC, F1 score, accuracy of algorithm. I it is greater than 0.60 that is sufficient, AUC – area under cover greater than 0.70 that is acceptable; and compare metrics of each algorithm and choose which algorithm metrics have good ROC, AUC, and accuracy. So accordingly two class-boosted decision algorithms fit for fertility diagnosis dataset.
![]()
![]()
Then I converted the training experiment into a predictive experiment, and after running the predictive experiment published the experiment as a web service.
![]()
The following is a screenshot of web service in that you will get API key, request response details for developer, ready made code in three languages C#, R and python. You can also test web service by clicking on the Test button.
![]()
![]()
The following is request response API documentation developer help page,
![]()