Graph Database as Catalyzer from Machine Learning in Pharma
1. Graph Database as a Catalyzer
from Machine Learning in
Pharma
Benito Gerónimo Marcos
2. Graph Databases
Graph databases are now clearly riding the upward trend toward
mainstream adoption for which the sector has been waiting for several
years.
Graph databases are utilized to explore relationships across massive
data and achieve analytics in real-time.
These applications can be used for artificial intelligence and machine
learning.
4. Graph Database in Pharma
• Medical Data is very heterogeneous.
• Pharma Data its behavior is huge
• Documents for expenses are fake
• Patient Response
• Churn Prediction
5. Machine Learning
• Machine Learning, by its nature, can be interact as a catalyzer per-se.
• Machine learning aid In pharma by means cleaning models for clean
and tagged data.
• sometimes records might be stored both in digital and physical
formats.
• and here is our proposal.
6. Graph Database as a Catalyzer
• Before Graph can be represent real-world
• Machine Learning can create models to clean and tagged data. As
well ML can be to predict through transformations.
• The question Could Graph Database turn into a Catalyzer for ML
models
• Answer: Yes
7. Graph Database as a Catalyzer
We take advantage of both emerging technologies Machine Learning
and Big Data through Graph Database.
we explore all of the relationship about medical data in order to create
a graph where we can look everything about the data and turn into
information and knowledge.
We utilized each node of the graph to represent a patient or a disease
or fake document or product for replenishing.
8. Graph Database as a Catalyzer
Transform all of data in information categorized and classified to
accelerate(Catalyzer) ML Models by means a new graph cleaned and
tagged.
9. Graph Database as a Catalyzer
Data
Physical
Data
Digital
Data
Cleaned
and
Tagged
ML Model
Edge
Classification
Data
11. Graph Database as a Catalyzer
• We take advantage of index adjacency for Neo4j and we take all of
nodes to classified and clustering data minimizing training time and
increasing performance for a Machine Learning.
12. Use Case
• Predict rare disease.
• Load medical data about diseases and regular behavior of illness patient
• Clean and tagged each node trough similarity search using each relationship
as a hamming distance.
• Our new graph obtain each node classified and weighted. This result is a new
input for a CNN model and finally predict if a new patient can get disease
• Some diseases predicted are diabetes, multiple sclerosis
• Predicted fake Documents for Claim fraud detection
• Churn Prediction