(MarketSummarySingleton, Log)
(MarketSummarySingleton, TradeConfig)
In this example embodiment, where the monolithic application comprises an example trading application, model component 108 can comprise a neural network such as, for instance, model 400 illustrated in FIG. 4. In this embodiment, microservice generation system 102 can train model component 108 (e.g., model 400) to learn the cluster assignments and/or graph embeddings described above by training model component 108 for the following task: given a specific class A in a runtime sequence, microservice generation system 102 can train model component 108 to predict the probability for every class of being the “neighboring” class of A in a runtime sequence. In this embodiment, if two Java classes have very similar “contexts,” it can mean these two classes are likely to co-occur within the same context window under the same business context and microservice generation system 102 can train model component 108 to output similar embedding for these two classes. In this embodiment that utilizes sequence-based embedding, sequences of neighboring nodes can be sampled (e.g., via microservice generation system 102 and/or model component 108) from the causal graph(s) using a method such as, for instance, random walk and the objective can be to minimize the negative log likelihood of observing the neighborhoods of nodes.
