Hands-On Machine Learning for Algorithmic Trading

Model evaluation

Basic functionality includes identifying similar words:

model.wv.most_similar(positive=['iphone'],                       restrict_vocab=15000)                 term  similarity0              android    0.6004541           smartphone    0.5816852                  app    0.559129

We can also validate individual analogies using positive and negative contributions accordingly:

model.wv.most_similar(positive=['france', 'london'],                       negative=['paris'],                       restrict_vocab=15000)             term  similarity0  united_kingdom    0.6066301         germany    0.5856442     netherlands    0.578868