Search

This work was supported by project PID2020-119478GB-I00 granted by Ministerio de Ciencia, Innovacion y Univesidades, project P18-FR-4961 by Proyectos I+D+i Junta de Andalucia 2018 and the process no 2015/20606-6, FundacAo de Amparo a Pesquisa do Estado de SAo Paulo (FAPESP) ., Data in the real world is far from being perfect. The appearance of noise is a common issue that arises from the limitations of data acquisition mechanisms and human knowledge. In classification, label noise will hinder the performance of almost all classifiers, inducing a bias in the built model. While label noise has recently attracted researchers’ attention in standard classification, it has only recently begun to be studied in multiple instance classification. In this work, we propose the usage of filtering algorithms for multiple instance classification that are able to reduce the impact of negative instances within the bags. In order to do so, we decompose the bags to form a standard classification problem that can be efficiently treated by a specialized noise filter. Such a decomposition is tackled in different ways, with the aim of exploiting the knowledge offered by the examples from opposite bags. The bags are then rebuilt, without the identified noise instances. In our experiments, we show that by applying our approach we can diminish the impact of noise and even obtain better results at 0% noise level for several classifiers. Our approach sets out a promising approach to dealing with noise in the bags of multiple instance datasets and further improve the classification rate of the built models., Ministerio de Ciencia, Innovacion y Univesidades PID2020-119478GB-I00, Junta de Andalucia P18-FR-4961, Fundacao de Amparo a Pesquisa do Estado de Sao Paulo (FAPESP) 2015/20606-6

Smart agriculture applications that analyse and manage agricultural yield using IoT systems may suffer from intermittent operation due to cloud disconnections commonly occurring in rural areas. A fog computing solution enables the IoT system to process data faster and deal with intermittent connectivity. However, the fog needs to send a high volume of data to the cloud and this can cause link congestion with unusable data traffic. Here we propose an approach to collect and store data in a fog-based smart agriculture environment and different data reduction methods. Sixteen techniques for data reduction are investigated; eight machine learning (ML) methods combined with run-length encoding, and eight combined with Huffman encoding. Our experiment uses two real data sets, where the first contains air temperature and humidity values, and the second has soil moisture and temperature conditions. The fog filters cluster the unlabelled data using unsupervised machine learning algorithms that group data into categories according to their value ranges in all experiments. Supervised learning classification methods are also used to predict the class of data samples from these categories. After that, the fog filter compresses the identified categories using two data compression techniques, run-length encoding (RLE) and the Huffman encoding, preserving the data time series nature. Our results reveal that a k-means combined with RLE method achieved the highest reduction, where the fog needed to store and transmit only 3%–6% of the original data generated by sensors.

The amount of quantum chemistry (QC) data is increasing year by year due to the continuous increase of computational power and development of new algorithms. However, in most cases, our atom-level knowledge of molecular systems has been obtained by manual data analyses based on selected descriptors. In this work, we introduce a data mining framework to accelerate the extraction of insights from QC datasets, which starts with a featurization process that converts atomic features into molecular properties (AtoMF). Then, it employs correlation coefficients (Pearson, Spearman, and Kendall) to investigate the AtoMF features relationship with a target property. We applied our framework to investigate three nanocluster systems, namely, PtnTM55-n, CenZr15-nO30, and (CHn + mH)/TM13. We found several interesting and consistent insights using Spearman and Kendall correlation coefficients, indicating that they are suitable for our approach; however, our results indicate that the Pearson coefficient is very sensitive to outliers and should not be used. Moreover, we highlight problems that can occur during this analysis and discuss how to handle them. Finally, we make available a new Python package that implements the proposed QC data mining framework, which can be used as is or modified to include new features.

The conceptualization of the musical timbre, which allows its quantitative evaluation in an audio record, is still an open-ended issue. This paper presents a set of dimensionless descriptors to assess the musical timbre of woodwind instruments in recordings of the fourth octave of the tempered musical scale. These descriptors are calculated from the Fast Fourier Transform (FFT) spectra using the Python Programming Language, specifically the SciPy library. The characteristic spectral signature of the clarinet, bassoon, transverse flute, and oboe are obtained in the fourth musical octave, observing the presence of degeneration for some musical sounds, that is, two given different aerophones may present the same harmonics. It is concluded that the proposed descriptors are sufficient to differentiate the aerophones studied, allowing their recognition, even in the case that there present the same set of harmonic frequencies.

Most machine learning applications in quantum-chemistry (QC) data sets rely on a single statistical error parameter such as the mean square error (MSE) to evaluate their performance. However, this approach has limitations or can even yield incorrect interpretations. Here, we report a systematic investigation of the two components of the MSE, i.e., the bias and variance, using the QM9 data set. To this end, we experiment with three descriptors, namely (i) symmetry functions (SF, with two-body and three-body functions), (ii) many-body tensor representation (MBTR, with two- and three-body terms), and (iii) smooth overlap of atomic positions (SOAP), to evaluate the prediction process's performance using different numbers of molecules in training samples and the effect of bias and variance on the final MSE. Overall, low sample sizes are related to higher MSE. Moreover, the bias component strongly influences the larger MSEs. Furthermore, there is little agreement among molecules with higher errors (outliers) across different descriptors. However, there is a high prevalence among the outliers intersection set and the convex hull volume of geometric coordinates (VCH). According to the obtained results with the distribution of MSE (and its components bias and variance) and the appearance of outliers, it is suggested to use ensembles of models with a low bias to minimize the MSE, more specifically when using a small number of molecules in the training set.

BackgroundPsoriatic arthritis (PsA) is an inflammatory arthritis associated with various rheumatological manifestations, such as arthritis or enthesitis, predominantly in the peripheral skeleton. However, the axial skeleton may be affected, as shown clinically or by conventional radiographs in up to 70% of patients with peripheral involvement. While there are studies comparing joint damage on standard radiography between patients with PsA and patients with ankylosing spondylitis [1], to our knowledge, no study has specifically evaluated the extent of structural lesions of the sacroiliac joints (SIJs) on computed tomography (CT) in patients with PsA compared with healthy controls.ObjectivesTo describe SIJ CT characteristics in patients with PsA and compare them with those of age- and sex-matched controls.MethodsAn observational, retrospective study was performed using medical records from Besançon and Dijon University Hospital’s rheumatology departments, which were screened to identify patients with PsA, according to the CASPAR criteria. A search was then carried out for patients in the hospitals’ imaging archiving system to identify those who had undergone a CT which included the SIJs in their entirety. Non-inclusion criteria were the existence of pelvic bone lesions and a history of pelvic radiotherapy. Each patient was then matched with a control of the same age and sex, recruited through the hospital’s imaging archiving system.For each individual, CT was interpreted by two independent readers using a score previously used by Diekhoff et al. [2], dividing each SIJ into 12 regions, for each of which joint space narrowing (JSN), erosions, and sclerosis are assessed. For this study, we also observed the existence of intra-articular gas and diffuse idiopathic skeletal hyperostosis (DISH) lesions for each region. Quantitative variables were compared using Student’s t-test. Qualitative variables were compared using the Chi-2 test.Results48 patients and 48 controls were included. Mean (SD) age was 54.76 ± 12.91 in the PsA group and 54.74 + 12.87 in the control group. 26 (54.18%) were male in both groups. In PsA patients, mean (SD) disease duration was 22.87 ± 14.95 years, 10 (43.48 %) were HLA-B27 positive, and 1 (2.86%) had a bamboo spine. CT findings are described in Table 1. The only lesion found significantly more frequently in PsA patients was erosion, which appeared to be preferentially located on the anterior and middle regions of the SIJs (Figure 1). A positive CT scan (significant joint space narrowing, erosion and/or sclerosis) was found in 15 (32.61%) of the patients with peripheral involvement and 6 (30.00%) of the patients with axial involvement.Table 1.Sacro-iliac CT findings using a score modified from Diekhoff et al.FindingPsA PatientsControlsp-valueMean (SD) total score (range 0-264)26.37 ± 29.1214.47 ± 10.850.01Global positivity, n (%)16 (33.33 %)10 (20.83 %)0.17Bilateral ankylosis, n (%)5 (10.42 %)0 (0.00 %)0.02Positive joint space score, n (%)15 (31.25 %)10 (20.83 %)0.25Positive erosion score, n (%)9 (18.75 %)1 (2.08 %)0.008Significant sclerosis, n (%)11 (22.92 %)12 (25.00 %)0.81Intra-articular gas, n (%)29 (60.42 %)35 (72.92 %)0.19DISH, n (%)9 (18.75%)11 (22.92 %)0.62Figure 1.Mean scores by region on anterior, middle, and posterior slices (JS: joint space, Ero: erosions, Scl: sclerosis) in PsA patients (A) and controls (B).ConclusionThe CT characteristics of SIJs from patients with PsA were similar to those of age- and sex-matched controls, but with a higher prevalence of erosions. Structural lesions of the SIJs were found in nearly one PsA patient out of three.References[1]Jadon DR, Sengupta R, Nightingale A, et al. Axial Disease in Psoriatic Arthritis study: defining the clinical and radiographic phenotype of psoriatic spondyloarthritis. Ann Rheum Dis. 2017;76:701–7.[2]Diekhoff T, Hermann K-GA, Greese J, et al. Comparison of MRI with radiography for detecting structural lesions of the sacroiliac joint using CT as standard of reference: results from the SIMACT study. Ann Rheum Dis. 2017;76(9):1502-1508.Disclosure of InterestsNone declared

BackgroundDifferentiating ankylosing spondylitis (AS) from osteitis condensans ilii (OCI) remains challenging for clinicians.ObjectivesThe aim of this study was to determine whether Subchondral Bone Attenuation Coefficient of the SacroIliac margins (SBAC-SI) is different in AS, OCI and diffuse idiopathic skeletal hyperostosis (DISH).MethodsA monocentric retrospective observational study was performed at a French University Hospital. Patients included were followed for AS, DISH or OCI and underwent CT scan including sacroiliac joint. Patients with tumor lesion of bone or a history of pelvic radiotherapy were excluded. AS and OCI patients were matched with a control of the same age and sex. All scans were acquired on the same CT-scan unit (Somatom 64 definition AS+, Siemens Healthineers, Erlangen, Germany), with a slice thickness of 0.625 mm. In the coronal oblique plane of the SIJ, three slices (anterior, middle and posterior) and four quadrants per joint were defined. Twenty-four identical circular regions of interest (ROIs) (30 mm2), 8 per slice, were manually placed separately subcortical to the SIJ, four on the sacral side and four on the iliac side. The distance between the circle of the ROI and the cortical bone was 2 to 3 mm. An overall score was obtained from the sum of all ROIs. For every ROI, an Attenuation Coefficient was measured and expressed in Hounsfield Unit. The total SBAC-SI score was the sum of the 24 ROI. The sacral and iliac SBAC-SI scores were the sum of the sacral or the iliac ROI.ResultsThirty AS and AS controls, 31 DISH, 29 OCI and OCI controls were included. SBAC-SI score was 9727 (±2430) in the OCI group (pConclusionSBAC-SI is significatively different between AS and OCI and could help to distinguish these two diseases.Figure 1.Distribution of the Subchondral Bone Attenuation Coefficients of the Sacroiliac Margins of the 24 ROI in the anterior (A), middle (B) and posterior (C) slices in the coronal oblique plane of the sacroiliac jointsDisclosure of InterestsNone declared

BackgroundOphthalmological manifestations in rheumatic diseases include dry syndrome, scleritis, episcleritis, uveitis and peripheral ulcerative keratitis (PUK). Among the target therapies, JAK inhibitors (JAKinhib) are the most recent to have been approved in rheumatic diseases, rheumatoid arthritis (RA) and spondyloarthritis (SpA), but also in other indications such as inflammatory bowel disease.ObjectivesThe aim of this work is to describe and characterise ophthalmological events in patients exposed to JAKinhib based on European pharmacovigilance (PV) data.MethodsThe ophthalmological manifestations that appeared under JAK inhib were extracted from the European PV database (April 2020), EUDRAVIGILANCE following a request addressed to the National Drug Safety Agency carried out according to the following MedDRA classification “SOC eye disorders and BARICITINIB” and “SOC eye disorders and TOFACITINIB”.ResultsA total of 1411 patients with ophthalmological adverse events (AEs) were identified with JAKinhib. 103 with BARICITINIB (81 women, mean age 59.7 years), 1308 with TOFACITINIB (1116 women, mean age 64.7 years). Among these AEs, 58% were reported by medical and paramedical staff and 42% were reported by patients. JAKinhib was initiated mainly in patients with RA (1070 patients), SpA (26 patients) and ulcerative colitis (UC) (27 patients). Of the reported AEs, 10% were inflammatory disorders of the anterior or posterior segments of the eye [51 uveitis (40 RA, 1 SpA, 1 RCUH, 1AJI, 8 undetermined (n.d.)), 49 corneal ulcerations (38 RA, 2 RCHU, 9 n.d.), 28 scleritis (20 RA, 8 n.d.), 13 keratitis (8 RA, 1 SpA, 1 Crohn’s, 3 n.d.)]. Visual loss and complete loss of vision were also reported in 200 patients. Finally, retinal detachment, retinal vascular thrombosis, cataracts and unspecified ophthalmological AEs were observed in 27, 25, 329 and 185 patients respectively. The mean time to onset of inflammatory ophthalmological events was 258 days after initiation of treatment (Table 1).Table 1.Ophthalmological adverse events with JAK inhibitorsBARICITINIBTOFACITINIBMean delay (in days)n=103n=1308Inflammatory involvement2 (2)26 (2)190±197 Scléritis, n (%)8 (8)41 (3)321±405 Corneal ulcération, n (%)1 (1)50 (4)292±296 Uvéitis, n (%)2 (2)11 (1)172±244 KératitisDecreased visual acuity, n (%)10 (10)118 (9)165±193Visual loss, n (%)072 (5)74±83Red eyes, n (%)14 (14)51 (4)80±149Dry eyes, n (%)15 (15)136 (10)166±454Photophobia, n (%)3 (3)14 (1)272±519Retinal detachment, n (%)1 (1)26 (2)461±360Retinal vascular thrombosis, n (%)3 (3)22 (2)367±490Retinal haemorrhage, n (%)2 (2)19 (1)168±199Cataract, n (%)9 (9)320 (24)Glaucoma, n (%)077 (6)432±360Age related macular degeneration (%)034 (3)351±549Unspecified ophthalmological AEs8 (8)177 (14)126±184ConclusionOphthalmological manifestations under JAK inhib seem rare but not exceptional. The rheumatologist must be made aware of them in order to discuss the potential imputability of the treatment and to report these manifestations to the pharmacovigilance structures. A detailed history, exclusion of infections and histopathological evaluation of the lesions are recommended to ensure that a differential diagnosis is not ignored. Topical treatments, and if necessary, discontinuation of the drug and switching to another targeted therapy may be considered. Discontinuation of JAKinhib appears to be warranted pending ophthalmologic advice.Disclosure of InterestsNone declared

BackgroundGrowing evidence indicated that the intestine is not only a target but also an actor of the pathogenesis in chronic inflammatory rheumatic diseases (CIRD). Consistently, increased intestinal permeability (IP) and damage (ID) as well as bacterial translocation (BT) have been described in patients with CIRD. However, the effects of treatments used in patients with CIRD on gut health are unknown.ObjectivesTo determine the effect of glucocorticoids (GCs) and tofacitinib on IP, ID and BT in rats with adjuvant-induced arthritis (AIA).MethodsAIA was induced in 6-week-old male Lewis rats by a tail injection of Mycobacterium butyricum in incomplete Freund’s adjuvant. At onset of arthritis, rats were treated daily with prednisolone at low (0.1 mg/kg/day, i.p.) or high dose (10 mg/kg/day, i.p.), or with Tofacitinib (10 mg/kg twice a day, s.c.) or with vehicle. After 21 days, IP, ID and BT were assessed by measurement of plasma levels of zonulin, intestinal Fatty Acid Binding Protein (iFABP) and serum levels of soluble CD14 (sCD14) by ELISA), respectively. Arthritis severity was daily evaluated through the determination of an arthritis score.ResultsCompared to vehicle, Tofacitinib and high-dose of GC both reduced arthritis score (pConclusionPrednisolone at a dose efficient on arthritis did not worsen but on the contrary reduced intestinal bacterial translocation and epithelial damage. These results are consistent with the positive effects of GCs on intestinal dysfunction observed in case of sepsis or colitis. The lack of efficacy of the sub-therapeutic dosage of prednisolone suggested that effects of GC are, at least partly, related to their anti-inflammatory effects. Consistent with the positive effect of jakinibs in patients with inflammatory bowel disease, Tofacitinib blunted intestinal bacterial translocation in AIA. Given the suspected pathophysiological link between bacterial translocation and arthritis, our results identified a new mechanism involved in the positive effects of GC and tofacitinib in arthritis diseases.Disclosure of InterestsNone declared

BackgroundThere is little epidemiological data on mortality in spondyloarthritis (SpA).ObjectivesTo determine countries’ mortality rates of ankylosing spondylitis (AS) and psoriatic arthritis (PsA), as well as chronic inflammatory bowel diseases (IBDs), which are physiologically related to SpA, and to describe their evolution between 2001 and 2015.MethodsWe used mortality data from the World Health Organisation (WHO), freely available on its website, which shows the number of deaths classified by age, sex, and cause of death coded by ICD-10. The code M45 was used for AS, L405 for PsA, K50 for Crohn’s disease (CD), and K51 for ulcerative colitis (UC). Age-standardized mortality rates (ASMR) were constructed using the 2015 WHO reference population and are expressed as deaths per million inhabitants. Temporal trend analyses of ASMR were performed between 2001 and 2015, considering only countries with up to 3 years of missing data, using joinpoint regression.ResultsIn 2015, the global ASMR of AS was 0.13 (0.11-0.14), ranging from 0.02 in Japan to 2.00 in Iceland (Figure 1A). The ASMR was 0.20 (0.18-0.23) for men and 0.07 (0.06-0.09) for women (pFigure 1.2015 ASMR by country for ankylosing spondylitis (A), psoriatic arthritis (B), Crohn’s disease (C) and ulcerative colitis (D).For PsA, the global ASMR in 2015 was 0.04 (0.03-0.05), ranging from 0.01 in Mexico to 0.13 in Greece (Figure 1B). The ASMR was 0.06 (0.04-0.07) for men and 0.03 (0.02-0.04) for women (p=0.01). The trend analysis showed a significant increase from 2004 to 2015 with a mean annual percent change (APC) of 5.94% (p=0.02). The ASMR in Europe (0.05 (0.03-0.06)) was significantly higher than in South America (0.02 (0.00-0.03)) (p=0.02).For CD, the global ASMR in 2015 was 0.86 (0.82-0.89), ranging from 0.02 in Thailand to 5.25 in Luxembourg (Figure 1C). The ASMR was 0.41 (0.38-0.43) for men and 0.38 (0.36-0.41) for women (p=0.17). The trend analysis showed a significant decrease from 2001 to 2015 with a mean APC of -0.82% (p=0.048). The ASMR in Europe (1.12 (1.05-1.19)) was significantly lower than in North America (1.31 (1.21-1.41)) (p=0.001), but significantly higher than in Latin America (0.57 (0.51-0.64)) (pFor UC, the global ASMR in 2015 was 0.76 (0.73-0.79), ranging from 0.03 in Thailand to 5.48 in Saint Lucia (Figure 1D). The ASMR was 0.37 (0.35-0.39) for men and 0.23 (0.21-0.24) for women (pConclusionASMR for IBD are higher than those for SpA and are decreasing over time, in contrast to SpA where they remain essentially stable. There are geographical disparities which must be interpreted with caution due to the declarative nature of the data.Disclosure of InterestsNone declared

BackgroundSacroiliac bone marrow edema is an important factor in the diagnosis and management of axial spondyloarthritis (axSpA).ObjectivesThe aim of this meta-analysis is to assess the effect of the different bDMARDs and tsDMARDs on the SPARCC score at 12-16 and 48-52 weeks.MethodsA systematic review, performed on Pubmed, Embase and Cochrane databases, included randomized controlled studies evaluating the sacroiliac joint (SIJ) SPARCC score at 12-16 or 48-52 weeks in patients with axSpA meeting the ASAS 2009 criteria or the modified New York criteria. We included studies evaluating the effects of the different treatments on the SPARCC score of SIJ in axial spondyloarthritis in comparison to a control group.ResultsFifteen studies were included in the meta-analysis. Nine studies evaluated the effect of TNF inhibitors (TNFi), two for ixekizumab, one for filgotinib and one for tofacitinib. At 12 and 16 weeks, SIJ SPARCC score was significantly improved by TNFi (WMD: - 3.29 [95% CI -4.25; -2, 34]), IL-17 inhibitors (WMD: - 3.98 [95% CI –5.76; -2.20]) and by JAK inhibitors (JAKi) (WMD: - 3.02 [95% CI -4.54; -1.51]). There was no difference between the molecule subgroups. At 48-52 weeks, TNFα inhibitors, introduced at baseline, reduced more SIJ SPARCC, but not significantly (WMD: -2.26 [95% CI -4.94; 0.42]), than placebo groups who began a TNFi treatment with delay.ConclusionOur meta-analysis shows a comparable improvement of the SIJ SPARCC score regarding TNFi, JAKi and IL-17 inhibitors at three months and suggests the presence of an opportunity window.Figure 1.Forest plot of SPARCC SIJ score for all therapeutic agents versus controls at week 12-16.Disclosure of InterestsNone declared

BackgroundPlaque Psoriasis (PsO) is an inflammatory disorder that could be associated with rheumatological manifestations defining Psoriatic arthritis. Moreover, recent studies showed that the presence of plaque psoriasis is associated with more structural damage In axial Spondyloarthritis [1]. To our knowledge, no study has specifically evaluated the extent of structural lesions of the sacroiliac joints (SIJs) on computed tomography (CT) in patients with PsO, without rheumatological manifestations, compared with healthy controls.ObjectivesTo describe SIJ CT characteristics in patients with PsO and compare them with those of age- and sex- matched controls.MethodsAn observational, retrospective study was performed using medical records from Besançon University Hospital’s rheumatology and dermatology departments, which were screened to identify patients with PsO, diagnosed by a dermatologist. A search was then carried out for patients in the hospitals’ imaging archiving system to identify those who had undergone a CT which included the SIJs in their entirety. Non-inclusion criteria were the existence of pelvic bone lesions and a history of pelvic radiotherapy. Each patient was then matched with a control of the same age and sex, recruited through the hospital’s imaging archiving system.For each individual, CT was interpreted by two independent readers using a score previously used by Diekhoff et al. [2], dividing each SIJ into 12 regions, for each of which joint space narrowing (JSN), erosions, and sclerosis are assessed. For this study, we also observed the existence of intra-articular gas and diffuse idiopathic skeletal hyperostosis (DISH) lesions for each region. Quantitative variables were compared using Student’s t-test. Qualitative variables were compared using the Chi-2 test.Results60 patients and 457controls were included. Mean (SD) age was 52,2 ± 17,7 in the PsO group and 53,6 ± 16.7 in the control group. 31 (52%) were male in PsO group and 30 (54%) in the control group. In the PsO group, BMI was 27,7± 6 versus 26,9 ± 6,44 in control group (p= 0.55). There is a trend for more frequent active smoking in PsO (69% in PsO group versus 54% in control group; p=0.075). In PsO patients, mean disease duration was 20.2 ± 17,6 years, the skin area affected was 41,8% ± 22,6 and the mean number of bDMARDs used was 2.25 (± 1.16). The global SIJ score was higher in the PsO group (6,63 ± 10,7) in comparison with the control group (2,84 ± 4,87). Erosion and sclerosis scores were similar in both group but joint space narrowing score was significantly higher in the PsO group (0.873 (±4.62) versus 4.15 (±10.8); p =0.035). There were no correlations between the global score and the disease duration (Pearson score 0.166 (-0.131; 0.435)) and the severity of the psoriasis (Pearson 0.00937 (-0.259; 0.276)).The number of gestation, active smoking, alcohol intake and physical work have no impact on the global Score.ConclusionThe CT characteristics of SIJs from patients with Pso were different to those of age- and sex-matched controls essentially secondary to joint space narrowing.References[1]Lorenzin M, Ortolan A, Vio S, Cozzi G, Scapin V, De Conti G, Doria A, Ramonda R. Psoriasis and Imaging Progression in early Axial Spondyloarthritis: Results from a 48-month follow-up (Italian arm of SPACE study). Rheumatology (Oxford). 2021 Oct 4:keab728.Table 1.CT scan findings between PsO patients and matched controlsControlsPsOErosion0.42 (±0.91)0.2 (±0.48)0.18Joint space narrowing0.873 (±4.62)4.15 (±10.8)0.035Sclerosis1.76 (±2.07)2.17 (±2.73)0.37Global score2.84 (±4.87)6.63 (±10.7)0.015Gas34 (62%)34 (57%)0.57DISH11 (20%)9 (15%)0.48Disclosure of InterestsNone declared

One of the significant challenges in machine learning is the classification of imbalanceddata. In many situations, standard classifiers cannot learn how to distinguish minority classexamples from the others. Since many real problems are unbalanced, this problem has becomevery relevant and deeply studied today. This paper presents a new preprocessing method based onDelaunay tessellation and the preprocessing algorithm SMOTE (Synthetic Minority Over-samplingTechnique), which we call DTO-SMOTE (Delaunay Tessellation Oversampling SMOTE). DTO-SMOTEconstructs a mesh of simplices (in this paper, we use tetrahedrons) for creating synthetic examples.We compare results with five preprocessing algorithms (GEOMETRIC-SMOTE, SVM-SMOTE,SMOTE-BORDERLINE-1, SMOTE-BORDERLINE-2, and SMOTE), eight classification algorithms,and 61 binary-class data sets. For some classifiers, DTO-SMOTE has higher performance than othersin terms of Area Under the ROC curve (AUC), Geometric Mean (GEO), and Generalized Index ofBalanced Accuracy (IBA).

[EN] Irony detection is a not trivial problem and can help to improve natural language processing tasks as sentiment analysis. When dealing with social media data in real scenarios, an important issue to address is data skew, i.e. the imbalance between available ironic and non-ironic samples available. In this work, the main objective is to address irony detection in Twitter considering various degrees of imbalanced distribution between classes. We rely on the emotIDM irony detection model. We evaluated it against both benchmark corpora and skewed Twitter datasets collected to simulate a realistic distribution of ironic tweets. We carry out a set of classification experiments aimed to determine the impact of class imbalance on detecting irony, and we evaluate the performance of irony detection when different scenarios are considered. We experiment with a set of classifiers applying class imbalance techniques to compensate class distribution. Our results indicate that by using such techniques, it is possible to improve the performance of irony detection in imbalanced class scenarios., The first author was funded by CONACYT project FC-2016/2410. Ronaldo Prati was supported by the São Paulo State (Brazil) research council FAPESP under project 2015/20606-6. Francisco Herrera was partially supported by the Spanish National Research Project TIN2017-89517-P. The work of Paolo Rosso was partially supported by the Spanish MICINN under the research project MISMIS (PGC2018-096212- B-C31) and by the Generalitat Valenciana under the grant PROMETEO/2019/121.