Modeling scientometric indicators using a statistical data ontology

Enhancing value creation of operational management for small to medium manufacturer: A conceptual data-driven analytical system

Purpose – This paper deals with the identification and diagnosis of operational variability in chemical processes, which is a common problem in mills but little explored in literature. The Cross-Industry Standard Process for Data Mining (CRISP-DM) is a widely used approach in problem solving. The purpose of this paper is to: first, contribute to the body of knowledge on applying CRISP-DM in a pulp mill production process and the special issues that need to be considered in this context. Exact amounts of a cost increase due to variation in pulp production have not been reported previously. Second, to quantify the cost of variation. Design/methodology/approach – In the case studied, the variation in a pulp mill batch cooking process had increased. In order to identify the causes of variation, CRISP-DM was applied. Findings – The cycle of variation was identified and found to be related to the batch cooking process cycle time. By using information from this analysis it was possible to detect otherwise unobserved defective steam nozzles. The defective equipment was repaired and improved. Further improvement was achieved when the fouling of a heat exchanger was found by analysis to be the root cause of long-term variability parameters. By applying CRISP-DM, equipment defects and fouling were identified as the root causes of the higher manufacturing costs due to increased variation were detected and estimated. The Taguchi loss function is a possible tool for estimating the cost of variation in pulp manufacturing. Originality/value – This paper provides new knowledge in the context of implementing CRISP-DM and the Taguchi loss function in the pulp and paper manufacturing process.

Football, which is a popular world-wide sport, has become one of the most practiced sports but also, with more study cases. Scouting and game analysis that is currently made has offered the possibility to improve the competition and increase the performance levels within a team. Taking this into account it emerged the term Scouting. The objective of this study is to streamline the Scouting process in Football, through Data Mining (DM) techniques and following the Cross Industry Standard Process for Data Mining (CRIPS-DM) methodology. The goal of DM was to develop and evaluate predictive models capable of forecasting a score of a football player’s performance. Based on this target, 2808 classification models and 936 regression models were developed and evaluated. For the classification, the maximum accuracy percentage was centered at 94% for the Forward player position, while for the regression the minimum error value was 0.07 for the Forward position. The results obtained allow to streamline the Scouting process in Football thus enhancing the sporting advantage.

Automatic detection of knee medial collateral ligament (MCL) tear from magnetic resonance imaging using deep neural network.

Cybercrime represents a growing global threat with adverse impacts on citizen security, the digital economy, and quality of life. In this context, an optimized hybrid model was developed that combines autoregressive integrated moving average (ARIMA) and long short-term memory (LSTM) for the monthly forecast of cybercrime complaints, applying the cross industry standard process for data mining (CRISP-DM) methodology and applying Python based data science techniques. The model combines the capabilities of the ARIMA statistical approach to capture linear components with the power of LSTM neural networks to address nonlinear temporal relationships. The architecture was trained on a set of 60,378 official records of complaints registered by the National Police of Peru between 2018 and 2023, achieving a mean absolute percentage error (MAPE) of 10.73%, which represents a significant improvement over the singular ARIMA and LSTM predictive models. Compared to previous studies in crime, health, and agriculture, this approach showed a greater ability to generalize over complex time series. It is concluded that the application of the proposed model is a relevant contribution for the police and other security agencies to anticipate crime trends and design preventive and effective strategies to combating cybercrime.

BackgroundCurrently, the diagnosis of Alzheimer’s disease dementia (ADD) is determined based on clinical criteria, as well as specific imaging and cerebrospinal fluid (CSF) biomarker profiles. However, healthcare professionals face a variety of challenges that hinder their application, such as the interpretation and integration or large amounts of data derived from neuropsychological assessment, the importance attributed to each source of information and the impact of unknown variables, among others. Therefore, this research focuses on the development of a computerized diagnostic tool based on Artificial Intelligence (AI), to strengthen the capacity of healthcare professionals in the identification and diagnosis of ADD.MethodDuring this research, all phases of the Cross‐Industry Standard Process for Data Mining (CRISP‐DM) methodology were applied. Specifically, seven types of machine learning (ML) classifier algorithms were trained with data derived from the medical and neuropsychological evaluation of 870 people with and without ADD. Additionally, genetic algorithms (GAs) were used to choose the hyperparameters of six of the models and the best performing model was deployed through a graphical user interface.ResultA total of 147 ML models were trained to differentiate or classify between people with and without ADD. Using a random forest (RF) algorithm and 31 predictor variables we achieved a sensitivity of 98%, a specificity of 100% and an area under the curve (AUC) of 0,988.ConclusionWe trained and evaluated a variety of ML models, highlighting the accuracy of a RF algorithm together with the use of GAs, to distinguish people with and without ADD. Additionally, we deployed the model with the best performance metrics through a graphical user interface in order to promote its future use in clinical settings.Although these results are promising, the ongoing need to evaluate and re‐train the model, as well as the ethical considerations of its implementation, are well acknowledged.

ObjectivesTo develop and to propose a machine learning model for predicting glaucoma and identifying its risk factors.MethodData analysis pipeline is designed for this study based on Cross-Industry Standard Process for Data Mining (CRISP-DM) methodology. The main steps of the pipeline include data sampling, preprocessing, classification and evaluation and validation. Data sampling for providing the training dataset was performed with balanced sampling based on over-sampling and under-sampling methods. Data preprocessing steps were missing value imputation and normalization. For classification step, several machine learning models were designed for predicting glaucoma including Decision Trees (DTs), K-Nearest Neighbors (K-NN), Support Vector Machines (SVM), Random Forests (RFs), Extra Trees (ETs) and Bagging Ensemble methods. Moreover, in the classification step, a novel stacking ensemble model is designed and proposed using the superior classifiers.ResultsThe data were from Shahroud Eye Cohort Study including demographic and ophthalmology data for 5190 participants aged 40-64 living in Shahroud, northeast Iran. The main variables considered in this dataset were 67 demographics, ophthalmologic, optometric, perimetry, and biometry features for 4561 people, including 4474 non-glaucoma participants and 87 glaucoma patients. Experimental results show that DTs and RFs trained based on under-sampling of the training dataset have superior performance for predicting glaucoma than the compared single classifiers and bagging ensemble methods with the average accuracy of 87.61 and 88.87, the sensitivity of 73.80 and 72.35, specificity of 87.88 and 89.10 and area under the curve (AUC) of 91.04 and 94.53, respectively. The proposed stacking ensemble has an average accuracy of 83.56, a sensitivity of 82.21, a specificity of 81.32, and an AUC of 88.54.ConclusionsIn this study, a machine learning model is proposed and developed to predict glaucoma disease among persons aged 40-64. Top predictors in this study considered features for discriminating and predicting non-glaucoma persons from glaucoma patients include the number of the visual field detect on perimetry, vertical cup to disk ratio, white to white diameter, systolic blood pressure, pupil barycenter on Y coordinate, age, and axial length.

The rapid growth of AI-generated content from ChatGPT, Gemini, and many others poses significant challenges in digital media. There are increasing instances of text generated by machine learning models being indistinguishable from human-written text. This calls for an effective way of detecting and distinguishing text generated by AI from human text. Using a Cross-Industry Standard Process for Data Mining (CRISP-DM) methodology, a Bidirectional Long Short-Term Memory (Bi-LSTM) model for classifying AI-generated text in a digital media context was developed. A carefully curated dataset of human and AI-generated text samples was used. The Bi-LSTM model without an embedding layer was implemented, optimizing the model to capture complex linguistic patterns apparent in each text type. An experimental setup was used to evaluate the effectiveness of the model. It was noted that the model achieved a remarkable test accuracy of 99.79%, with a loss of 0.009. To facilitate practical implementation, we developed a web application using Next.js with our model served from a Flask server that enables real-time AI text detection. Our results highlight the model’s ability to accurately identify AI-generated text, providing valuable insights into deploying such models to verify content on media platforms. This study highlights the potential of neural network-based classifiers to address the pressing need for automated AI text detection in an increasingly AI-influenced digital ecosystem.

This study aims to implement a classification method using the Nave Bayes Classifier (NBC) algorithm on Lake Toba visitor review text data. The Cross Industry Standard Process for Data Mining (CRISP-DM) methodology comprises the following stages: business understanding, data understanding, data preparation, modeling, evaluation, and deployment. The findings of this study indicate that during the phase of business comprehension, the context of the discussion focuses on the tourism sector, specifically tourist perceptions of the quality of products and services at Lake Toba tourist destinations. At the data comprehension stage, the source of review data used was the Tripadvisor website, which contained as many as 858 reviews with the following rating classification: 8 reviews with abysmal ratings; 22 reviews with poor ratings; 81 reviews with neutral ratings; 304 reviews with good ratings; 443 reviews with excellent ratings. Data cleansing is performed at the data preparation stage so that 382 data are processed by dividing training data by 70 percent and test data by 30 percent. During the modeling phase, the performance of the NBC and DT algorithms was evaluated using and without SMOTE UPsampling operators. The comparison of NBC and DT algorithm values indicates that the model with the best performance is DT using SMOTE UPsampling operators with accuracy values (98.27 percent), precision values (98.83 percent), recall values (97.71 percent), f-measure values (98.26 percent), and AUC values (98.27 percent) (0.982). At the evaluation stage, the importance of excellent service (Quality Human Resources) and supporting infrastructure was highlighted by analyzing the results of ranking the five most frequently used terms in Lake Toba visitor review data (tourism facilities and infrastructure). At the deployment stage, it is necessary to balance the development of attractions, accessibility, lodging, and tourism-supporting amenities to generate visiting intention and revisit motivation to Lake Toba.

Loss of blood circulation to the brain causes a stroke, which is also known as a brain attack. In addition, blood clots are also the leading cause of stroke in the brain. Based on the WHO report, stroke is the leading cause of death in Indonesia in 2024, with a death rate reaching 131.8 per 100,000 population. This study aims to classify early detection of stroke disease by applying the Cross Industry Standard Process for Data Mining (CRISP-DM) methodology using the Random Forest algorithm. The data used is public through the website www.kaggle.com, with a total of 4981 records consisting of 11 attributes. The data composition is unbalanced, with 4733 negative stroke data (95%) and 248 positive strokes (5%). Handling imbalanced data using the Synthetic Minority Oversampling Technique (SMOTE). The total data from SMOTE is 5981 records, with 4733 negative data and 1248 positive. After exploring several models, the best model was obtained using Random Forest with the SMOTE approach, producing an accuracy of 80.14%, AUC 0.836, recall 63.33%, and precision 11.42%.

Outpatient clinics are integral to healthcare, offering vital services without the need for hospital admission. However, appointment scheduling in these settings remains challenging due to uncertainties such as patient no-shows and variable service times. This study proposes a data-driven approach to minimize physician idle time and patient waiting time by analyzing an eight-year (2016-2023) dataset from primary and specialized care for American veterans. Adopting the Cross Industry Standard Process for Data Mining (CRISP-DM) methodology, four predictive models: Random Forest, Extreme Gradient Boosting (XGBoost), Adaptive Boosting (AdaBoost), and Artificial Neural Networks (ANN), were developed for both classification (noshow) and regression (service time) tasks. The ANN model demonstrated superior predictive performance in both domains. The key predictors of no-shows included waiting time, type of care and care provider, while type of care, care provider, and veteran ZIP code were the most influential in forecasting service time. These findings highlight the potential of machine learning to improve appointment scheduling in outpatient clinics.

One of the key performance indicators for the logistics industry, especially freight forwarder company (cargo), is the delivery time. This is still a challenge in this industry in terms of ensuring the customer service level and reducing transportation costs. On the other hand, the development of information technology now allows an organization or company to collect large amounts of data automatically. A decent method that can be used to analyze the data for prediction purposes is machine learning, which is a method of extracting data into a certain pattern of information. This research aims to apply three machine learning methods to estimate the status of shipping goods. The method used in this study follows the machine learning process published by the Cross Industry Standard Process for Data Mining (CRISP-DM), namely; business processes understanding, data understanding, data preparation, model development, evaluation, and implementation. Based on the results of the study, the random forest method produces better accuracy than the logistic regression and artificial neural network (ANN) methods, which is 76.6%, while the results of ANN and logistic regression methods are 73.81% and 72.84% respectively.

This research aims to explore tourist behaviour patterns and develop data-driven tourism route recommendations based on Association Rule Mining (ARM), particularly using the Frequent Pattern Growth (FP-Growth) algorithm. The study focuses on identifying co-occurrence patterns among tourist attractions based on real behavioral data, with the goal of supporting tourism route design and enhancing recommendations for self-guided tourists. The study employed a structured data mining methodology. Raw data was collected from real-world tourist check-ins and GPS traces in Songkhla City, Thailand; encompassing 45 attraction points. The study used the Cross-Industry Standard Process for Data Mining (CRISP-DM) for research methodology. The FP-Growth algorithm was used to discover association rules, using parameters set at support > 0.10, confidence > 0.5, lift > 1.0, and conviction > 1.0 to ensure the relevance and reliability of the extracted rules. The evaluation phase included interpretation of rule quality metrics such as support, confidence, lift, and conviction indicators. Visualizations presented the route recommendations on geographic maps. The results revealed 13 valid association rules involving only 5 out of the 45 locations, suggesting strong behavioral patterns and repeat co-visitation. Interestingly, some of the top-visited locations based on frequency alone did not appear in any of the 13 rules, indicating that frequency-based route planning may not reflect actual tourist behavior. Further analysis showed that all route combinations extracted from the rules had a total travel distance of under 88 kilometers. With Google Map data, this study demonstrates that ARM with FP-Growth can provide insightful patterns for route planning and destination clustering. The proposed method supports smart tourism development by bridging behavioral analytics and operational tourism planning. Stakeholders can apply this approach to periodically update route recommendations to ensure alignment with current tourist behaviors and preferences.

This study aims to predict the location of theft crimes in the jurisdiction of the Badung Police by applying the K-Nearest Neighbor (KNN) method. The main focus of the study is to identify crime patterns based on time and location variables in order to improve the effectiveness of police prevention strategies. The problem raised is how machine learning-based models can help detect theft-prone areas and improve accuracy in crime prevention efforts. This study uses a quantitative approach with the CRISP-DM (Cross-Industry Standard Process for Data Mining) method. The data used includes information on time, location of the incident, and theft categories based on police reports. The research process includes business understanding, data exploration and preparation, modeling using KNN, model performance evaluation, and implementation in the form of interactive map visualization. Model performance is analyzed using evaluation metrics such as precision, recall, and F1-score to measure the level of prediction accuracy. The results of the study show that the KNN model is able to identify locations with a high risk of theft with fairly good accuracy. Areas with high activity, such as transportation facilities and commercial areas, are more vulnerable to this crime. In addition, thefts occur more often in the morning, evening, and early morning when people are off guard. In conclusion, the KNN method is effective in predicting theft-prone areas. Implementation of this model can help the police improve the effectiveness of patrols and security strategies. It is recommended that this model be combined with a geographic information system (GIS) to facilitate the analysis of crime patterns in order to improve public security more proactively.

This study targets service-delivery improvements in South African public hospitals by pairing Big Data analytics with machine learning, implemented under a quantitative Cross-Industry Standard Process for Data Mining (CRISP-DM) workflow. We (i) identified determinants of effective Big Data analytics in public healthcare, (ii) compared analytical methods, (iii) built a predictive model of service outcomes, and (iv) assessed its performance and interpretability. Using a synthetic hospital dataset, we engineered features and trained a stacked ensemble (XGBoost, Random Forest, CatBoost, LightGBM) to classify outcomes as Improved, Unchanged, or Worsened. Baseline models (e.g., Support Vector Machine (SVM), logistic regression) were explored but not retained due to scalability and multiclass performance constraints. The final ensemble achieved 84% accuracy with class-wise Area Under the Curves (AUCs) of 0.93-0.95. Post-hoc SHapley Additive exPlanations (SHAP) analysis highlighted Satisfaction Rating, Age, and log (Wait Time) as the most influential features, and revealed interaction effects such as Wait-length interactions that align with operational intuition. To support practical use, we prototyped a Streamlit application that delivers near-real-time, explainable predictions for front-line decision support. The findings indicate that Machine Learning (ML)-driven Big Data analytics can provide scalable, interpretable insights to inform predictive decision-making, resource allocation, and patient-flow management in the public sector. While results are demonstrated on synthetic data, the approach establishes a deployable pathway for hospital analytics in South Africa and a foundation for validation on real-world datasets.