Violation prediction apparatus, violation prediction method and program

A violation prediction method executed by a violation prediction apparatus that predicts an occurrence probability of a violation using a memory and a processor coupled to the memory, the violation prediction method comprising:

• storing conversation data and relationship data in the memory, the conversation data including conversation content data and conversation situation data, the relationship data indicating a relationship with a target user;
• receiving the conversation data and the relationship data by the processor from the memory;
• evaluating the conversation data and identifying a partner of a conversation by performing machine-implemented extraction of a conversation content and a conversation situation from the conversation data, thereby obtaining a conversation content evaluation value and a conversation situation evaluation value;
• evaluating a relationship with the partner of the conversation based on the relationship data and conversation partner information, thereby obtaining a relationship evaluation value;
• calculating a social interaction effect based on the conversation content evaluation value, the conversation situation evaluation value, and the relationship evaluation value;
• calculating a time attenuation value of the social interaction effect based on a time attenuation function executed by the processor;
• calculating a violation level of a target action;
• training a probability-prediction model using machine-learning training techniques based on the violation level of the target action and the social interaction effect subjected to time attenuation processing;
• and
• calculating an occurrence probability of a violation based on the social interaction effect subjected to time attenuation processing by executing the trained probability-prediction model.