top of page
Predictive system for medical program inclusion based on doctor notations
CLIENT
Medical Institution
TEAM
Skilled data scientists and software developers who possessed extensive knowledge and expertise in NLP, machine learning, and computer vision
INDUSTRY
Healthcare
TECHNOLOGIES
Python, Numpy, Pandas, Yolov5, Pose-Net, Torch, Tensorflow, Tf-Lite (for transfer model to Nvidia Jetson), Sklearn, Nltk.
About the project
Our client wanted to develop a predictive system that could analyze doctors' notations and determine whether a patient should be included or excluded from a medical program. Additionally, they wanted to implement machine learning techniques to predict a patient's readmission during the 30-day period.
Challenge
The challenge was that the doctor notations were in text format, making it difficult to extract relevant information. Our solution was built on NLP system using Bert models that could transform the textual data into numerical vectors and identify relevant features for inclusion in the predictive model.
Solution
After conducting exploratory data analysis, our team of experts utilized a range of machine learning techniques, including Logistic Regression, Random Forest, Naive Bayes, and Bagging, to develop a classification model that could accurately predict a patient's eligibility for a medical program based on doctor notations.
We also optimized the pipeline and cleaned the code to ensure efficient and accurate predictions. To further enhance the system's accuracy, we integrated Yolov5, Pose-Net, Torch, Tensorflow, and Tf-Lite (for transfer model to Nvidia Jetson) to improve the prediction capabilities.
Impact
Upon completion of the project the client
01
got the access to an advanced accounting system and internet shop engine
02
benefited from streamlined financial processes, improved data analysis, and enhanced retail operations
03
The integration of AI and ML technologies provided valuable insights, optimized decision-making, and boosted overall efficiency.
Featured projects
bottom of page