ideas for ML training

[jexp]

• wrongly categorized businesses
• determine businesses who changed owners
• predict score a user would give to a business based on both of their histories together
• review fraud can we teach a model to determine fraudulent reviews based on structure?
• review patterns per category (distribution)
• clustering with deepgl
• KNN
• textrank or reviews?

[mneedham]

• Collaborative filtering

[mneedham]

• Jaccard similarity?

[mneedham]

• Label propagation - show how to do the semi supervised variant with pre defined labels. Could be based on an amenity similarity graph?

[mneedham]

• Feeding Graph Algo output into an ML prediction model?

[mneedham]

• Content based filtering based on amenities?

[mneedham]

• Similar properties based on embeddings?