Application Of Machine Learning Methods Research Articles

An Adaptive Technique for Crime Rate Prediction using Machine Learning Algorithms

Any country must give the investigation and preventive of crime top priority. There are a rising amount of cases that are still pending due to the rapid increase in criminal cases in India and elsewhere. It is proving difficult to classify and address the rising number of criminal cases. Understanding a place's trends in criminal activity is essential to preventing it from occurring. Crime-solving organisations will be more effective if they have a clear awareness of the patterns of criminal behavior that are present in a particular area. Women's safety and protection are of highest importance despite the serious and persistent problem of crime against them. This study offers predictions about the kinds of crimes that might occur in a particular location using ensemble methods. This facilitates the categorization of criminal proceedings and subsequent action in a timely manner. We are applying machine learning methods like KNN, Linear regression, SVM, Lasso, Decision tree and Random forest in order to assess the highest accuracy.

Generating a Single Session Outcome Measure from Digital Mental Health Platform Footprints Using Natural Language Processing

Generating a Single Session Outcome Measure from Digital Mental Health Platform Footprints Using Natural Language Processing

Aqueous dissolution of Li-Na borosilicates: Insights from machine learning and experiments

Previously acquired data could be utilised in predicting glass dissolution kinetics at long times, but the application of machine learning methods needs to be assessed. Here, the dissolution processes of two Li-Na borosilicate ‘base glasses’ at 40 and 90 °C were investigated by SEM-EDS, NMR and Raman spectroscopy. Boron and sodium machine learning predictions were excellent when considering other normalised releases as features. However, extrapolating the training feature space yielded poorer performance and the absence of incorporated waste elements resulted in underestimated predicted long-term lithium and silicon releases. Faster dissolution kinetics were observed for MW than MW-½Li but the MW-½Li gel layer at 40 °C trapped more water. Whilst BO3 rings leached preferentially at 90 °C, surface enrichment of BO3 at 40 °C suggested [BO4]− transformed prior to dissolution. Results were consistent with interdiffusion being significant at 40 °C and interface-coupled dissolution precipitation beyond 7 days at 90 °C.

Impact of salinity gradient, water pollution and land use types on greenhouse gas emissions from an urbanized estuary

Estuaries have been recognized as one of the major sources of greenhouse gases (GHGs) in aquatic systems; yet we still lack insights into the impact of both anthropogenic and natural factors on the dynamics of GHG emissions. Here, we assessed the spatiotemporal dynamics and underlying drivers of the GHG emissions from the Scheldt Estuary with a focus on the effects of salinity gradient, water pollution, and land use types, together with their interaction. Overall, we found a negative impact of salinity on carbon dioxide (CO2) and nitrous oxide (N2O) emissions which can be due to the decrease of both salinity and water quality when moving upstream. Stronger impact of water pollution on the GHG emissions was found at the freshwater sites upstream compared to saline sites downstream. In particular, when water quality of the sites reduced from good, mainly located in the mouth and surrounded by arable sites, to polluted, mainly located in the upstream and surrounded by urban sites, CO2 emissions from the sites doubled while N2O emissions tripled. Similarly, the effects of water pollution on methane (CH4) emissions became much stronger in the freshwater sites compared to the saline sites. These decreasing effects from upstream to the mouth were associated with the increase in urbanization as sites surrounded by urban areas released on average almost two times more CO2 and N2O than sites surrounded by nature and industry areas. Applied machine learning methods also revealed that, in addition to salinity effects, nutrient and organic enrichment stimulated the GHG emissions from the Scheldt Estuary. These findings highlight the importance of the interaction between salinity, water pollution, and land use in order to understand their influences on GHG emissions from dynamic estuarine systems.

Machine Learning Study of SNPs in Noncoding Regions to Predict Non-small Cell Lung Cancer Susceptibility

Non-small cell lung cancer (NSCLC) is the most common pathological subtype of lung cancer. Both environmental and genetic factors have been reported to impact the lung cancer susceptibility. We conducted a genome-wide association study (GWAS) of 287 NSCLC patients and 467 healthy controls in a Chinese population using the Illumina Genome-Wide Asian Screening Array Chip on 712,095 SNPs (single nucleotide polymorphisms). Using logistic regression modeling, GWAS identified 17 new noncoding region SNP loci associated with the NSCLC risk, and the top three (rs80040741, rs9568547, rs6010259) were under a stringent p-value (<3.02e-6). Notably, rs80040741 and rs6010259 were annotated from the intron regions of MUC3A and MLC1, respectively. Together with another five SNPs previously reported in Chinese NSCLC patients and another four covariates (e.g., smoking status, age, low dose CT screening, sex), a predictive model by machine learning methods can separate the NSCLC from healthy controls with an accuracy of 86%. This is the first time to apply machine learning method in predicting the NSCLC susceptibility using both genetic and clinical characteristics. Our findings will provide a promising method in NSCLC early diagnosis and improve our understanding of applying machine learning methods in precision medicine.

Handling Ill-conditioned Omics Data with Deep Probabilistic Models.

The advent of high-throughput technologies has produced an increase in the dimensionality of omics datasets, which limits the application of machine learning methods due to the great unbalance between the number of observations and features. In this scenario, dimensionality reduction is essential to extract the relevant information within these datasets and project it in a low-dimensional space, and probabilistic latent space models are becoming popular given their capability to capture the underlying structure of the data as well as the uncertainty in the information. This article aims to provide a general classification and dimensionality reduction method based on deep latent space models that tackles two of the main problems that arise in omics datasets: the presence of missing data and the limited number of observations against the number of features. We propose a semi-supervised Bayesian latent space model that infers a low-dimensional embedding driven by the target label: the Deep Bayesian Logistic Regression (DBLR) model. During inference, the model also learns a global vector of weights that allows it to make predictions given the low-dimensional embedding of the observations. Since this kind of dataset is prone to overfitting, we introduce an additional probabilistic regularization method based on the semi-supervised nature of the model. We compared the performance of the DBLR against several state-of-the-art methods for dimensionality reduction, both in synthetic and real datasets with different data types. The proposed model provides more informative low-dimensional representations, outperforms the baseline methods in classification, and can naturally handle missing entries.

A Computed Tomography-based Radiomics Analysis of Low-energy Proximal Femur Fractures in the Elderly Patients.

Low-energy proximal femur fractures in elderly patients result from factors, like osteoporosis and falls. These fractures impose high rates of economic and social costs. In this study, we aimed to build predictive models by applying machine learning (ML) methods on radiomics features to predict low-energy proximal femur fractures. Computed tomography scans of 40 patients (mean ± standard deviation of age = 71 ± 6) with low-energy proximal femur fractures (before a fracture occurs) and 40 individuals (mean ± standard deviation of age = 73 ± 7) as a control group were included. The regions of interest, including neck, trochanteric, and intertrochanteric, were drawn manually. The combinations of 25 classification methods and 8 feature selection methods were applied to radiomics features extracted from ROIs. Accuracy and the area under the receiver operator characteristic curve (AUC) were used to assess ML models' performance. AUC and accuracy values ranged from 0.408 to 1 and 0.697 to 1, respectively. Three classification methods, including multilayer perceptron (MLP), sequential minimal optimization (SMO), and stochastic gradient descent (SGD), in combination with the feature selection method, SVM attribute evaluation (SAE), exhibited the highest performance in the neck (AUC = 0.999, 0.971 and 0.971, respectively; accuracy = 0.988, 0.988, and 0.988, respectively) and the trochanteric (AUC = 1, 1 and 1, respectively; accuracy = 1, 1 and 1, respectively) regions. The same methods demonstrated the highest performance for the combination of the 3 ROIs' features (AUC = 1, 1 and 1, respectively; accuracy =1, 1 and 1, respectively). In the intertrochanteric region, the combination methods, MLP + SAE, SMO + SAE, and SGD + SAE, as well as the combination of the SAE method and logistic regression (LR) classification method exhibited the highest performance (AUC = 1, 1, 1 and 1, respectively; accuracy= 1, 1, 1 and 1, respectively). Applying machine learning methods to radiomics features is a powerful tool to predict low-energy proximal femur fractures. The results of this study can be verified by conducting more research on bigger datasets.

A Comprehensive Prediction Method for Pore Pressure in Abnormally High-Pressure Blocks Based on Machine Learning

In recent years, there has been significant research and practical application of machine learning methods for predicting reservoir pore pressure. However, these studies frequently concentrate solely on reservoir blocks exhibiting normal-pressure conditions. Currently, there exists a scarcity of research addressing the prediction of pore pressure within reservoir blocks characterized by abnormally high pressures. In light of this, the present paper introduces a machine learning-based approach to predict pore pressure within reservoir blocks exhibiting abnormally high pressures. The methodology is demonstrated using the X block as a case study. Initially, the combination of the density–sonic velocity crossplot and the Bowers method is favored for elucidating the overpressure-to-compact mechanism within the X block. The elevated pressure within the lower reservoir is primarily attributed to the pressure generated during hydrocarbon formation. The Bowers method has been chosen to forecast the pore pressure in well X-1. Upon comparison with real pore pressure data, the prediction error is found to be under 5%, thus establishing it as a representative measure of the reservoir’s pore pressure. Intelligent prediction models for pore pressure were developed using the KNN, Extra Trees, Random Forest, and LightGBM algorithms. The models utilized five categories of well logging data, sonic time difference (DT), gamma ray (GR), density (ZDEN), neutron porosity (CNCF), and well diameter (CAL), as input. After training and comparison, the results demonstrate that the LightGBM model exhibits significantly superior performance compared to the other models. Specifically, it achieves R2 values of 0.935 and 0.647 on the training and test sets, respectively. The LightGBM model is employed to predict the pore pressure of two wells neighboring well X-1. Subsequently, the predicted data are juxtaposed with the actual pore pressure measurements to conduct error analysis. The achieved prediction accuracy exceeds 90%. This study delivers a comprehensive analysis of pore pressure prediction within sections exhibiting anomalously high pressure, consequently furnishing scientific insights to facilitate both secure and efficient drilling operations within the X block.

Soil Salinity Estimation for South Kazakhstan Based on SAR Sentinel-1 and Landsat-8,9 OLI Data with Machine Learning Models

Climate change, uneven distribution of water resources and anthropogenic impact have led to salinization and land degradation in the southern regions of Kazakhstan. Identification of saline lands and their mapping is a laborious process associated with a complex of ground measurements. Data from remote sensing are widely used to solve this problem. In this paper, the problem of assessing the salinity of the lands of the South Kazakhstan region using remote sensing data is considered. The aim of the study is to analyze the applicability of machine learning methods to assess the salinity of agricultural lands in southern Kazakhstan based on remote sensing. The authors present a salinity dataset obtained from field studies and containing more than 200 laboratory measurements of soil salinity. Moreover, the authors describe the results of applying several regression reconstruction algorithms (XGBoost, LightGBM, random forest, Support vector machines, Elastic net, etc.), where synthetic aperture radar (SAR) data from the Sentinel-1 satellite and optical data in the form of spectral salinity indices are used as input data. The obtained results show that, in general, these input data can be used to estimate salinity of the wetted arable land. XGBoost regressor (R2 = 0.282) showed the best results. Supplementing the radar data with the values of salinity spectral index improves the result significantly (R2 = 0.356). For the local datasets, the best result shown by the model is R2 = 0.473 (SAR) and R2 = 0.654 (SAR with spectral indexes), respectively. The study also revealed a number of problems that justify the need for a broader range of ground surveys and consideration of multi-year factors affecting soil salinity. Key results of the article: (i) a set of salinity data for different geographical zones of southern Kazakhstan is presented for the first time; (ii) a method is proposed for determining soil salinity on the basis of synthetic aperture radar supplemented with optical data, and this resulted in the improved prediction of the results for the region under consideration; (iii) a comparison of several types of machine learning models was made and it was found that boosted models give, on average, the best prediction result; (iv) a method for optimizing the number of model input parameters using explainable machine learning is proposed; (v) it is shown that the results obtained in this work are in better agreement with ground-based measurements of electrical conductivity than the results of the previously proposed global model.

Reputational assets and social media marketing activeness: Empirical insights from China

We explore the linkages between social media marketing activeness and reputational assets on digital platforms with a unique sample of over 8,000 customer-to-customer (C2C) sellers registered on both Taobao, China’s largest C2C online shopping platform, and Sina Weibo, China’s largest microblogging platform. A unique collaborative effort between the two platforms enables us to examine whether C2C sellers are motivated to engage in marketing activities on a separate social media platform. Applying machine learning methods, we first classify whether C2C sellers conduct social media marketing on their microblogs or not, which allows the measurement of social media marketing activeness. We then use logistic regression models and find that earned reputational assets such as the rating scores and the number of followers are significantly associated with social media marketing activeness on both platforms. However, we identify a conflict of owned reputational assets such as the shop age and the paid membership between the two platforms, which provides a potential explanation for the limited success of the cross-platform collaboration.

Addressing annotation and data scarcity when designing machine learning strategies for neurophotonics.

Machine learning has revolutionized the way data are processed, allowing information to be extracted in a fraction of the time it would take an expert. In the field of neurophotonics, machine learning approaches are used to automatically detect and classify features of interest in complex images. One of the key challenges in applying machine learning methods to the field of neurophotonics is the scarcity of available data and the complexity associated with labeling them, which can limit the performance of data-driven algorithms. We present an overview of various strategies, such as weakly supervised learning, active learning, and domain adaptation that can be used to address the problem of labeled data scarcity in neurophotonics. We provide a comprehensive overview of the strengths and limitations of each approach and discuss their potential applications to bioimaging datasets. In addition, we highlight how different strategies can be combined to increase model performance on those datasets. The approaches we describe can help to improve the accessibility of machine learning-based analysis with limited number of annotated images for training and can enable researchers to extract more meaningful insights from small datasets.

Seismic fragility analysis of steel moment frames using machine learning models

This study develops machine learning (ML) models for seismic fragility analysis of steel moment frames. Four ML methods – random forest, adaptive boosting, gradient boosting regression tree (GBRT), and extreme gradient boosting (XGBoost) – were employed for this purpose. Probabilistic seismic demand models, each representing the relationship between the seismic response of a type of structure and ground motion intensity, were used to construct the fragility curves based on nonlinear time history analyses of 616 steel moment frames subjected to 240 ground motions. The first three natural periods of steel moment frames and a capacity limit state defined by the maximum interstory drift were selected as input variables for the ML models. Two parameters (median and logarithmic standard deviation) of a fragility function were considered as output variables for the ML models. For each steel frame, the capacity limit state values considered for maximum interstory drift cover a wide range to generalize the fragility curve outcomes. The interquartile range method was used to ensure the quality of the dataset, and consequently 56,479 data points were used for the development of ML models. Based on model performance, the GBRT (R2 = 0.9986, for the testing dataset) and XGBoost (R2 = 0.9987) models are proposed as the best models for fragility analysis of steel moment frames. Finally, a graphical user interface for fragility analysis of steel moment frames was built based on the two proposed models, for easy access by practicing engineers. This study demonstrates the applicability of ML methods in practical design.

Serum metabolomics detected by LDI-TOF-MS can be used to distinguish between diabetic patients with and without diabetic kidney disease.

Diabetic kidney disease (DKD) is an important cause of end-stage renal disease with changes in metabolic characteristics. The objective of this study was to study changes in serum metabolic characteristics in patients with DKD and to examine metabolite panels to distinguish DKD from diabetes with matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF-MS). We recruited 40 type II diabetes mellitus (T2DM) patients with or without DKD from a single center for a cross-sectional study. Serum metabolic profiling was performed with MALDI-TOF-MS using a vertical silicon nanowire array. Differential metabolites between DKD and diabetes patients were selected, and their relevance to the clinical parameters of DKD was analyzed. We applied machine learning methods to the differential metabolite panels to distinguish DKD patients from diabetes patients. Twenty-four differential serum metabolites between DKD patients and diabetes patients were identified, which were mainly enriched in butyrate metabolism, TCA cycle, and alanine, aspartate, and glutamate metabolism. Among the metabolites, l-kynurenine was positively correlated with urinary microalbumin, urinary microalbumin/creatinine ratio (UACR), creatinine, and urea nitrogen content. l-Serine, pimelic acid, 5-methylfuran-2-carboxylic acid, 4-methylbenzaldehyde, and dihydrouracil were associated with the estimated glomerular filtration rate (eGFR). The panel of differential metabolites could be used to distinguish between DKD and diabetes patients with an AUC value reaching 0.9899-0.9949. Among the differential metabolites, l-kynurenine was related to the progression of DKD. The differential metabolites exhibited excellent performance at distinguishing between DKD and diabetes. This study provides a novel direction for metabolomics-based clinical detection of DKD.

Applying machine learning methods to predict operative mortality after tricuspid valve surgery

BackgroundEuroSCORE stratifies surgical risk in cardiac surgery; however, it is not explicitly for tricuspid valve surgery. Therefore, we aimed to apply machine learning (ML) methods to predict operative mortality after tricuspid valve surgery and compare the predictive ability of these models to EuroSCORE.This retrospective analysis included 1161 consecutive patients who underwent tricuspid valve surgery at a single center from 2009 to 2021. The study outcome was operative mortality (n=112), defined as mortality occurring within 30 days of surgery or the same hospital admission. Random forest, LASSO, elastic net, and logistic regression were used to identify predictors of operative mortality.ResultsEuroSCORE was significantly higher in patients who had operative mortality [8.52 (4.745–20.035) vs.4.11 (2.29–6.995), P<0.001] [AUC=0.73]. Random forest identified eight variables predicting operative mortality with an accuracy of 92% in the test set (age≥70 years, heart failure, emergency surgery, chronic kidney disease grade IV, diabetes mellitus, tricuspid valve replacement, hypertension, and redo surgery). The classification error rate in the training data was 9%, and in the testing data, it was 4.8%. Logistic regression identified eight variables with an AUC of 0.76. LASSO identified 13 variables with an AUC of 0.78, and elastic net identified 17 variables (AUC=0.795). The AUCs of the elastic net (P=0.048) and random forest (P<0.001) models were significantly higher than that of EuroSCORE.ConclusionsML effectively predicted TV surgery mortality more accurately than the traditional risk-scoring method. Incorporating ML in cardiac surgery risk scoring with comprehensive inclusion of all possible variables is recommended.

Large-scale reference-free analysis of flavivirus sequences in Aedes aegypti whole genome DNA sequencing data

Flaviviruses are a diverse group of RNA viruses, which include the etiological agents of Zika, dengue and yellow fever that are transmitted by mosquitoes. Flaviviruses do not encode reverse transcriptase and cannot reverse transcribe into DNA, yet DNA sequences of flaviviruses are found both integrated in the chromosomes of Aedes aegypti mosquitoes and as extrachromosomal sequences. We have previously examined the Ae. aegypti reference genome to identify flavivirus integrations and analyzed conservation of these sequences among whole-genome data of 464 Ae. aegypti collected across 10 countries globally. Here, we extended this analysis by identifying flavivirus sequences in these samples independently of the Ae. aegypti reference assembly. Our aim was to identify the complete set of viral sequences, including those absent in the reference genome, and their geographical distribution. We compared the identified sequences using BLASTn and applied machine learning methods to identify clusters of similar sequences. Apart from clusters of sequences that correspond to the four viral integration events that we had previously described, we identified 19 smaller clusters. The only cluster with a strong geographic association consisted of Cell-fusing agent virus-like sequences specific to Thailand. The remaining clusters did not have a geographic association and mostly consisted of near identical short sequences without strong similarity to any known flaviviral genomes. The short read sequencing data did not permit us to determine whether identified sequences were extrachromosomal or integrated into Ae. aegypti chromosomes. Our results suggest that Liverpool strain and field Ae. aegypti mosquitoes have a similar variety of conserved flaviviral DNA, whose functional role should be investigated in follow-up studies.Graphical

Machine learning in marine ecology: an overview of techniques and applications

AbstractMachine learning covers a large set of algorithms that can be trained to identify patterns in data. Thanks to the increase in the amount of data and computing power available, it has become pervasive across scientific disciplines. We first highlight why machine learning is needed in marine ecology. Then we provide a quick primer on machine learning techniques and vocabulary. We built a database of ∼1000 publications that implement such techniques to analyse marine ecology data. For various data types (images, optical spectra, acoustics, omics, geolocations, biogeochemical profiles, and satellite imagery), we present a historical perspective on applications that proved influential, can serve as templates for new work, or represent the diversity of approaches. Then, we illustrate how machine learning can be used to better understand ecological systems, by combining various sources of marine data. Through this coverage of the literature, we demonstrate an increase in the proportion of marine ecology studies that use machine learning, the pervasiveness of images as a data source, the dominance of machine learning for classification-type problems, and a shift towards deep learning for all data types. This overview is meant to guide researchers who wish to apply machine learning methods to their marine datasets.

Applying machine learning methods to quantify emotional experience in installation art

Aesthetic experience is original, dynamic and ever-changing. This article covers three research questions (RQs) concerning how immersive installation artworks can elicit emotions that may contribute to their popularity. Based on Yayoi Kusama’s and Peter Kogler’s kaleidoscopic rooms, this study aims to predict the emotions of visitors of immersive installation art based on their Twitter activity. As indicators, we employed the total number of likes, comments, retweets, followers, followings, the average of tweets per user, and emotional response. According to our evaluation of emotions, panic obtained the highest scores. Furthermore, compared to traditional machine learning algorithms, Tree-based Pipeline Optimization Tool (TPOT) Automated Machine Learning used in this research yielded slightly lower performance. We forecast that our findings will stimulate future research in the fields of data analysis, cultural heritage management and marketing, aesthetics and cultural analytics.

Machine learning for prediction of asthma exacerbations among asthmatic patients: a systematic review and meta-analysis

BackgroundAsthma exacerbations reduce the patient’s quality of life and are also responsible for significant disease burdens and economic costs. Machine learning (ML)-based prediction models have been increasingly developed to predict asthma exacerbations in recent years. This systematic review and meta-analysis aimed to identify the prediction performance of ML-based prediction models for asthma exacerbations and address the uncertainty of whether modern ML methods could become an alternative option to predict asthma exacerbations.MethodsPubMed, Cochrane Library, EMBASE, and Web of Science were searched for studies published up to December 15, 2022. Studies that applied ML methods to develop prediction models for asthma exacerbations among asthmatic patients older than five years and were published in English were eligible. The prediction model risk of bias assessment tool (PROBAST) was utilized to estimate the risk of bias and the applicability of included studies. Stata software (version 15.0) was used for the random effects meta-analysis of performance measures. Subgroup analyses stratified by ML methods, sample size, age groups, and outcome definitions were conducted.ResultsEleven studies, including 23 prediction models, were identified. Most of the studies were published in recent three years. Logistic regression, boosting, and random forest were the most used ML methods. The most common important predictors were systemic steroid use, short-acting beta2-agonists, emergency department visit, age, and exacerbation history. The overall pooled area under the curve of the receiver operating characteristics (AUROC) of 11 studies (23 prediction models) was 0.80 (95% CI 0.77–0.83). Subgroup analysis based on different ML models showed that boosting method achieved the best performance, with an overall pooled AUROC of 0.84 (95% CI 0.81–0.87).ConclusionThis study identified that ML was the potential tool to achieve great performance in predicting asthma exacerbations. However, the methodology within these models was heterogeneous. Future studies should focus on improving the generalization ability and practicability, thus driving the application of these models in clinical practice.

An active learning framework and assessment of inter-annotator agreement facilitate automated recogniser development for vocalisations of a rare species, the southern black-throated finch (Poephila cincta cincta)

The application of machine learning methods has led to major advances in the development of automated recognisers used to analyse bioacoustics data. To further improve the performance of automated call recognisers, we investigated the development of efficient data annotation strategies and how best to address uncertainty around ambiguous vocalisations. These challenges present a particular problem for species whose vocalisations are rare in field recordings, where collecting enough training data can be problematic and a species' vocalisations may be poorly documented.We provide an open access solution to address these challenges using two strategies. First, we applied an active learning framework to iteratively improve a convolutional neural network (CNN) model able to automate call identification for a target rare bird species, the southern black-throated finch (Poephila cincta cincta). We collected 9098 h of unlabelled audio recordings from a field study in the Desert Uplands Bioregion of Queensland, Australia, and used active learning to prioritise human annotation effort towards data that would best improve model fit. Second, we progressed methods for managing ambiguous vocalisations by applying machine learning methods more commonly used in medical image analysis and natural language processing. Specifically, we assessed agreement among human annotators and the CNN model (i.e. inter-annotator agreement) and used this to determine realistic performance outcomes for the CNN model and to identify areas where inter-annotator agreement may be improved. We also applied a classification approach that allowed the CNN model to classify sounds into an ‘uncertain’ category, which replicated a requirement of human-annotation and facilitated the comparison of human-model annotation performance.We found that active learning was an efficient strategy to build a CNN model where there was limited labelled training data available, and target calls were extremely rare in the unlabelled data. As few as five active learning iterations, generating a final labelled dataset of 1073 target calls and 5786 non-target sounds, were required to train a model to identify the target species with comparable performance to experts in the field.Assessment of inter-annotator agreement identified a bias in our model to align predictions most closely with those of the primary annotator and identified significant differences in inter-annotator agreement among subsets of our acoustic data. Our results highlight the use of inter-annotator agreement to understand model performance and identify areas for improvement in data annotation. We also show that excluding ambiguous vocalisations during data annotation results in an overestimation of model performance, an important consideration for datasets with inter-annotator disagreement.

Applying machine learning methods for the analysis of two-dimensional mass spectra

In a measurement of isomeric yield-ratios in fission, the Phase-Imaging Ion-Cyclotron-Resonance technique, which projects the radial motions of ions in the Penning trap (JYFLTRAP) onto a position-sensitive micro-channel plate detector, has been applied. To obtain the yield ratio, that is the relative population of two states of an isomer pair, a novel analysis procedure has been developed to determine the number of detected ions in each state, as well as corrections for the detector efficiency and decay losses. In order to determine the population of the states in cases where their mass difference is too small to reach full separation, a Bayesian Gaussian Mixture model was implemented. The position-dependent efficiency of the micro-channel plate detector was calibrated by mapping it with ^{133}Cs^+ ions, and a Gaussian Process was trained with the position data to construct an efficiency function that could be used to correct the recorded distributions. The obtained numbers of counts of excited and ground-state ions were used to derive the isomeric yield ratio, taking into account decay losses as well as feeding from precursors.