The Badas peat swamp forest, Brunei Darussalam, Borneo, is a large Baram river peatland formation facing pressures from natural and increasingly anthropogenic disturbances. Although the peat forest should regenerate naturally after disturbances, there is little quantitative information documenting this regeneration. Given the impenetrability and remoteness of peatland forests, remote sensing technology can fill the peatland forest regeneration knowledge gap. This paper reports on the findings of a study on the resilience of the Badas peat swamp forest, composed primarily of Shorea albida Sym. We identified the scars caused by logging and caterpillar infestation using photo interpretation and the forest inventory map. We employed the Shuttle Radar Topography Mission (SRTM), TanDEM-X 90 m (TDX) digital elevation products, and AW3D30 m (AW) digital surface model (DSM) to assess the regeneration of disturbed forest scars over 14 years. Through measurements of forest height and other indices, we found that the Badas peat swamp forest did not recover from these past disturbances and the depletion of scars progressed even further. Our findings support other authors’ conclusions suggesting irrevocable degradation of Bornean peatland and peat forests. We also introduce an accuracy assessment of the forest biomass estimation relying on a comparison of DSMs and present a detailed morphology of the surface of the Badas peat dome.

Leaf is one of the most commonly used organs for species identification. The traditional identification process involves a manual analysis of individual dried or fresh leaf’s features by the botanists. Recent advancements in computer vision techniques have assisted in automating the plants families/species identification process based on the digital images of leaves. However, most of the existing studies have focused on using datasets for fresh and intact leaves. A huge amount of data for preserved plants in the form of digitized herbaria specimens have not been effectively utilized for the task of automated identification because of the presence of damaged leaves in specimens. In this study, deep learning techniques have been proposed as a tool for reconstructing the damaged herbarium leaves in order to maximize the usefulness of the digitized specimens for automated plant identification task by increasing the number of individual samples of leaves. The reconstruction results of two different families of convolution neural networks (CNNs) have been compared for data from ten different plant families namely Anacardiaceae, Annonaceae, Dipterocarpaceae, Ebenaceae, Euphorbiaceae, Malvaceae, Phyllanthaceae, Polygalaceae, Rubiaceae and Sapotaceae. The performance of automated identification task was improved by more than 20% using the reconstructed leaves images as compared to using the original data (i.e. images of specimens with damaged leaves). This work evidently suggests that deep learning techniques can be utilized for reconstruction of damaged leaves even on a challenging herbarium leaves dataset.

Approximate frequent itemsets (AFI) mining from noisy databases are computationally more expensive than traditional frequent itemset mining. This is because the AFI mining algorithms generate large number of candidate itemsets. This article proposes an algorithm to mine AFIs using pattern growth approach. The major contribution of the proposed approach is it mines core patterns and examines approximate conditions of candidate AFIs directly with single phase and two full scans of database. Related algorithms apply Apriori-based candidate generation and test approach and require multiple phases to obtain complete AFIs. First phase generates core patterns, and second phase examines approximate conditions of core patterns. Specifically, the article proposes novel techniques that how to map transactions on approximate FP-tree, and how to mine AFIs from the conditional patterns of approximate FP-tree. The approximate FP-tree maps transactions on shared branches when the transactions share a similar set of items. This reduces the size of databases and helps to efficiently compute the approximate conditions of candidate itemsets. We compare the performance of our algorithm with the state of the art AFI mining algorithms on benchmark databases. The experiments are analyzed by comparing the processing time of algorithms and scalability of algorithms on varying database size and transaction length. The results show pattern growth approach mines AFIs in less processing time than related Apriori-based algorithms.

This work aims at developing a generalized and optimized path loss model that considers rural, suburban, urban, and urban high rise environments over different frequencies, for use in the Heterogenous Ultra Dense Networks in 5G. Five different machine learning algorithms were tested on four combined datasets, with a sum of 12369 samples in which their hyper-parameters were automatically optimized using Bayesian optimization, HyperBand and Asynchronous Successive Halving (ASHA). For the Bayesian optimization, three surrogate models (the Gaussian Process, Tree Structured Parzen Estimator and Random Forest) were considered. To the best of our knowledge, few works have been found on automatic hyper-parameter optimization for path loss prediction and none of the works used the aforementioned optimization techniques. Differentiation among the various environments was achieved by the assignment of the clutter height values based on International Telecommunication Union Recommendation (ITU-R) P.452-16. We also included the elevation of the transmitting antenna position as a feature so as to capture its effect on path loss. The best machine learning model observed is K Nearest Neighbor (KNN), achieving mean Coefficient of Determination (R2), average Mean Absolute Error (MAE) and mean Root Mean Squared Error (RMSE) values of 0.7713, 4.8860dB, and 6.8944dB, respectively, obtained from 100 different samplings of train set and test set. Results show that machine learning can also be used to develop path loss models that are valid for a certain range of distances, frequencies, antenna heights, and environment types. HyperBand produced hyperparameter configurations with the highest accuracy in most of the algorithms.

Road accidents occur at traffic signals, even though traffic patrolling and awareness are increased. Accidents occurred at traffic signal junctions are a significant proportion in the overall reported road accidents. Usually, when the drivers approaching the traffic signals, at the onset of yellow, the drivers would enter into a dilemma zone, where they will be in confusion mode assessing their capabilities to cross the intersection or stop. So, any improper decision might lead to a collision. To avoid right-angle crash, drivers apply harsh brake to stop before the traffic signals. But this may lead to back-end crash, when the following driver encounters the former’s sudden stopping decision. This situation gets multifaceted when the traffic is heterogeneous containing various types of vehicles. So, the main objective of this study is to assess the performance using machine learning techniques. In this study, support vector machine (SVM) and K-nearest neighbors’ (KNN) techniques are implemented to validate the classification of the driving behaviour in terms of safe stopping/unsafe stopping at the signalized junctions at the onset of the yellow signal.

Previously, cluster-based multi or many objective function techniques were proposed to reduce the Pareto set. Recently, researchers proposed such techniques to find better solutions in the objective space to solve engineering problems. In this work, we applied a cluster-based approach for solution selection in a multiobjective evolutionary algorithm based on decomposition with bare bones particle swarm optimization for data clustering and investigated its clustering performance. In our previous work, we found that MOEA/D with BBPSO performed the best on 10 datasets. Here, we extend this work applying a cluster-based approach tested on 13 UCI datasets. We compared with six multiobjective evolutionary clustering algorithms from the existing literature and ten from our previous work. The proposed technique was found to perform well on datasets highly overlapping clusters, such as CMC and Sonar. So far, we found only one work that used cluster-based MOEA for clustering data, the hierarchical topology multiobjective clustering algorithm. All other cluster-based MOEA found were used to solve other problems that are not data clustering problems. By clustering Pareto solutions and evaluating new candidates against the found cluster representatives, local search is introduced in the solution selection process within the objective space, which can be effective on datasets with highly overlapping clusters. This is an added layer of search control in the objective space. The results are found to be promising, prompting different areas of future research which are discussed, including the study of its effects with an increasing number of clusters as well as with other objective functions.