Evaluation of the Known Sub-Sequence Algorithm (KSSA) for Optimal Use in Time Series of Tuna Fishing in the Pacific Ocean

An important limitation in fisheries research using time series is the presence of missing data. An inadequate handling of these data can negatively impact the results of statistical analyses, leading to erroneous decision-making. A potential solution to this problem is the estimation of missing values through imputation methods, but none of them can be universally applied to all time series. In fact, their effectiveness largely depends on the structure of the data and the distribution of the missing values. Recently, a solution to this problem was proposed which employs the known sub-sequence algorithm (KSSA), a machine learning tehnique designed to compare the performance of different imputation methods and validate them within the time series that contains missing data. However, due to its recent development, there is no published evidence regarding its efficiency and reliability. This research aimed to assess the efficiency of the KSSA on tuna fisheries data for the Pacific Ocean by imputing simulated missing data for seven time series with distinct structures. The results demonstrated that the algorithm is robust for accurately validating a wide combination of properties, such as length, seasonality, trend, autocorrelation structure, and percentage of missing data. Additionally, the algorithm’s hyperparameters can be easily adjusted to achieve optimal results for each time series.