To collect information about the location and water depth of the fishing grounds, we attached data loggers to otter boards of three bottom otter trawlers in Ise Bay, with GPS data loggers on deck. We likewise asked the fishermen to record the outcomes of every haul on a logbook in chronological order. In the event that there are missing record(s) on the logbook, all succeeding records will be misaligned. As such, we will be unable to integrate the data with environmental/GPS data logger records. Thus, a method to recover the correspondence relationship between both records is needed to make the most of the acquired data.
Our method consists of two steps. First, we used a random forest model to calculate the probability that any combination of a logbook record and a data logger record are derived from the same haul. The differences of “latitude”, “longitude”, “heading”, “speed”, and “haul duration” between both records were used as explanatory variables. As for all possible combinations between the logbook records and data loggers, a maximum likelihood algorithm was applied to each record on the logbook. This could predict which data logger records are most likely to be integrated.
We examined the performance of the model by removing some records from the actual complete logbook records. If the missing record was once in a trip, the success rate of integration was very high. Without using “heading” as an explanatory variable, the success rate reduced drastically. The model was also applied to actual incomplete logbook records.