Predictive Analytics
The current process is a very manual and expensive process to determine whether a patient needs prescription or not, after analyzing blood work and scanning reports. Determination is based on multiple data sets and relies on extensive human judgment. The process is slow as a doctor needs to consider scanned data, Laboratory Reference Ranges by experts, and any pre-existing disease or treatment. Thus, the cost is expensive to determine if specific prescription is needed or not. With the increasing number of patients having secondary disease being ill, the hospital is not only losing the opportunity to treat increased number of patients by performing manual judgement but most importantly putting the patients’ lives at further risk due to high probability of human error.
Hybrid Model
As we realize the need of Predictive Analytics, Saimatics Solutions successfully experiments a hybrid model consists of two data streams, one relying on traditional marketing campaigns and the other on generalizable categorical features of new ad campaigns being fed into the world famous XG Boost machine learning model. We used a hybrid modeling program to run few tests of the campaigns and produced the confusion matrices and AUC scores to combine probabilities of the two models for further analysis.
Human Feedback to ML Prediction
In order to test accuracy of the hybrid model in production environment, we implemented a feedback mechanism within the live data streaming. So that sales reps can easily and consistently report back what is and isn’t working, and also comment on any other feature values improvements.
Conclusion
ML based Predictive Analysis will not only allow the companies to become data driven disruptor but also helps them to remain ahead of the competition in the new digital market.