Tree-based Machine Learning Models Research Articles

Robustness of Optimized Decision Tree-Based Machine Learning Models to Map Gully Erosion Vulnerability

Gully erosion is a worldwide threat with numerous environmental, social, and economic impacts. The purpose of this research is to evaluate the performance and robustness of six machine learning ensemble models based on the decision tree principle: Random Forest (RF), C5.0, XGBoost, treebag, Gradient Boosting Machines (GBMs) and Adaboost, in order to map and predict gully erosion-prone areas in a semi-arid mountain context. The first step was to prepare the inventory data, which consisted of 217 gully points. This database was then randomly subdivided into five percentages of Train/Test (50/50, 60/40, 70/30, 80/20, and 90/10) to assess the stability and robustness of the models. Furthermore, 17 geo-environmental variables were used as potential controlling factors, and several metrics were examined to evaluate the performance of the six models. The results revealed that all of the models used performed well in terms of predicting vulnerability to gully erosion. The C5.0 and RF models had the best prediction performance (AUC = 90.8 and AUC = 90.1, respectively). However, according to the random subdivisions of the database, these models exhibit small but noticeable instability, with high performance for the 80/20% and 70/30% subdivisions. This demonstrates the significance of database refining and the need to test various splitting data in order to ensure efficient and reliable output results.

Using machine learning to predict 1,2,3-trichloropropane contamination from legacy non-point source pollution of groundwater in California's Central Valley

1,2,3-tricholoropropane (TCP) is an impurity common in nematicides applied to agricultural soils from the 1940s to the 1980s. Evidence from animal studies indicates that TCP is a probable human carcinogen. TCP leaches through the soil into groundwater where it persists and contaminates thousands of wells in Asia, Europe, and North America. In California, TCP contaminates drinking water wells, with the highest levels of TCP beneath agricultural land used to grow grapevines. This study performs a mass balance and evaluates the ability of three types of tree-based machine learning models to predict TCP concentration in California's Central Valley aquifer system: classification and regression trees (CART), Random Forest (RF), and Boosted Regression Trees (BRT). To construct the models, multiple spatial explanatory variables were used, including historical agricultural land use data, irrigation levels, precipitation, soil type, groundwater age, redox state, and the presence of the co-contaminant nitrate. To estimate the amount of TCP applied to farmland in California, state historical pesticide and land use documents were used in the mass balance. Between 110,000 and 4,300,000 kg of TCP are estimated to have accumulated in the subsurface. Machine learning models indicate that the most important explanatory variables to predict TCP contamination of groundwater are precipitation, redox state, and the presence of the co-contaminant nitrate. Additionally, a 1000-m buffer area offers a slightly higher predictive performance as compared to 500-m and 1500-m buffers. Furthermore, the RF model outperforms CART and BRT for predictive performance. Finally, modeling using decision trees can predict TCP contamination levels in areas where monitoring is lacking, help target future TCP monitoring efforts, and aid in identifying areas to avoid when drilling new drinking water wells.

Machine-learning-based prediction and optimization of emerging contaminants' adsorption capacity on biochar materials

Biochar materials have recently received considerable recognition as eco-friendly and cost-effective adsorbents capable of effectively removing hazardous emerging contaminants (e.g., pharmaceuticals, herbicides, and fungicides) to aquatic organisms and human health accumulated in aquatic ecosystems. In this study, ten tree-based machine learning (ML) models, including bagging, CatBoost, ExtraTrees, HistGradientBoosting, XGBoost, GradientBoosting, DecisionTree, Random Forest, Light gradient Boosting, and KNearest Neighbors, have been built to accurately predict the adsorption capacity of biochar materials toward ECs in aqueous solutions. A very large data set with 3,757 data points was generated using 24 input variables (i.e., pyrolysis conditions for biochar production (3 features), biochar characteristics (3 features), biochar compositions (6 features), and adsorption experimental conditions (12 features)) obtained from the batch adsorption experiments to remove 12 kinds of ECs using 18 different biochar materials. The rigorous evaluation and comparison of the ML model performances shows that CatBoost model had the highest test coefficient of determination (0.9433) and lowest mean absolute error (4.95 mg/g), outperformed clearly all other models. The feature importance analyzed by the shapley additive explanations (SHAP) indicated that the adsorption experimental conditions provided the highest impact on the model prediction for adsorption capacity (41 %) followed by the adsorbent composition (35 %), adsorbent characterization (20 %), and synthesis conditions (3)%). The optimized experimental conditions predicted by the modeling were a N/C ratio of 0.017, BET surface area of 1040 m2/g, content of C(%) contents of 82.1 %, pore volume of 0.46 cm3/g, initial ECs concentration of 100 mg/L, type of pollutant (CAR), adsorption type (Single) and adsorption contact time (720 min).

Predictability of Belgian residential real estate rents using tree-based ML models and IML techniques

PurposeThe purpose is twofold. First, this study aims to establish that black box tree-based machine learning (ML) models have better predictive performance than a standard linear regression (LR) hedonic model for rent prediction. Second, it shows the added value of analyzing tree-based ML models with interpretable machine learning (IML) techniques.Design/methodology/approachData on Belgian residential rental properties were collected. Tree-based ML models, random forest regression and eXtreme gradient boosting regression were applied to derive rent prediction models to compare predictive performance with a LR model. Interpretations of the tree-based models regarding important factors in predicting rent were made using SHapley Additive exPlanations (SHAP) feature importance (FI) plots and SHAP summary plots.FindingsResults indicate that tree-based models perform better than a LR model for Belgian residential rent prediction. The SHAP FI plots agree that asking price, cadastral income, surface livable, number of bedrooms, number of bathrooms and variables measuring the proximity to points of interest are dominant predictors. The direction of relationships between rent and its factors is determined with SHAP summary plots. In addition to linear relationships, it emerges that nonlinear relationships exist.Originality/valueRent prediction using ML is relatively less studied than house price prediction. In addition, studying prediction models using IML techniques is relatively new in real estate economics. Moreover, to the best of the authors’ knowledge, this study is the first to derive insights of driving determinants of predicted rents from SHAP FI and SHAP summary plots.

Tree-based machine learning models assisted fluorescent sensor array for detection of metal ions based on silver nanocluster probe

The growing concern over heavy metal pollution and its impact on the environment and human health has led to a proliferation of research on the detection and differentiation of heavy metal ions. A novel fluorescent sensor array utilizing only one single Ag-nanoclusters (Ag NCs) was developed for the efficient detection of six metal ions. The Ag NCs probe was prepared by using poly(methyl vinyl ether-alt-maleic acid) (PMVEM) as the ligand and has different fluorescence properties in water and dimethyl sulfoxide (DMSO). The interaction between metal ions and Ag NCs resulted in a characteristic fluorescence variation pattern which was subsequently analyzed using various tree-based machine learning models. We have compared different combinations of classification models and pre-processing methods of which the K-Nearest Neighbors Classifier with the first five linear discriminants has the highest accuracy. Through the integration of concentration models within a tree-based pipeline optimization framework, six unique concentration regression models were selected for each metal ion. In addition, the developed sensor array could identify metal ions in binary mixtures. And it still kept high accuracy for the classification of six target metal ions in river water. In conclusion, the proposed framework was found to be effective in the detection of heavy metal ions in environmental samples, thus providing a promising approach for addressing heavy metal pollution.

A hybrid ensemble machine learning model for discharge coefficient prediction of side orifices with different shapes

Side orifices are widely applied for flow control and regulation in channel systems. Accurate estimation of the discharge coefficient of the side orifice is significant for water management. The main objective of current research is to accurately predict the discharge coefficients of circular and rectangular side orifices. Considering that traditional empirical regressions are hard to estimate the discharge coefficient precisely due to the complex nonlinear relationship between the discharge coefficient and relevant parameters, a new hybrid boosting ensemble machine learning model, BO-XGBoost, is developed, which combines the advantages of the boosting ensemble model (XGBoost) and Bayesian Optimization. To further evaluate the proposed hybrid model, it is also compared with other tree-based machine learning models, including standalone XGBoost, Random Forest (RF) and Decision Tree (DT). Literature experimental data of the flow and geometric parameters relevant to the discharge coefficients of circular and rectangular side orifices are collected and applied to develop the models. Four dimensionless parameters of the relative channel width (B/L), the relative bottom height (W/L), the relative upstream depth (Y/L) and the upstream Froude number (Fr) are taken into consideration for the prediction of discharge coefficient (Cd). Furthermore, four different input combinations are designed and then compared to determine the best one on the basis of RMSE. By using the optimal input combination, our results demonstrate that BO-XGBoost provides the best comprehensive performance among all the involved machine learning models in the discharge coefficient prediction for both types of side orifices. Besides, the uncertainty analysis also reveals that BO-XGBoost shows the narrowest uncertainty bandwidth and gives the highest prediction reliability.

Descriptor engineering in machine learning regression of electronic structure properties for 2D materials

We build new material descriptors to predict the band gap and the work function of 2D materials by tree-based machine-learning models. The descriptor’s construction is based on vectorizing property matrices and on empirical property function, leading to mixing features that require low-resource computations. Combined with database-based features, the mixing features significantly improve the training and prediction of the models. We find R^{2} greater than 0.9 and mean absolute errors (MAE) smaller than 0.23 eV both for the training and prediction. The highest R^{2} of 0.95, 0.98 and the smallest MAE of 0.16 eV and 0.10 eV were obtained by using extreme gradient boosting for the bandgap and work-function predictions, respectively. These metrics were greatly improved as compared to those of database features-based predictions. We also find that the hybrid features slightly reduce the overfitting despite a small scale of the dataset. The relevance of the descriptor-based method was assessed by predicting and comparing the electronic properties of several 2D materials belonging to new classes (oxides, nitrides, carbides) with those of conventional computations. Our work provides a guideline to efficiently engineer descriptors by using vectorized property matrices and hybrid features for predicting 2D materials properties via ensemble models.

Predicting road flooding risk with crowdsourced reports and fine-grained traffic data

The objective of this study is to predict road flooding risks based on topographic, hydrologic, and temporal precipitation features using machine learning models. Existing road inundation studies either lack empirical data for model validations or focus mainly on road inundation exposure assessment based on flood maps. This study addresses this limitation by using crowdsourced and fine-grained traffic data as an indicator of road inundation, and topographic, hydrologic, and temporal precipitation features as predictor variables. Two tree-based machine learning models (random forest and AdaBoost) were then tested and trained for predicting road inundations in the contexts of 2017 Hurricane Harvey and 2019 Tropical Storm Imelda in Harris County, Texas. The findings from Hurricane Harvey indicate that precipitation is the most important feature for predicting road inundation susceptibility, and that topographic features are more critical than hydrologic features for predicting road inundations in both storm cases. The random forest and AdaBoost models had relatively high AUC scores (0.860 and 0.810 for Harvey respectively and 0.790 and 0.720 for Imelda respectively) with the random forest model performing better in both cases. The random forest model showed stable performance for Harvey, while varying significantly for Imelda. This study advances the emerging field of smart flood resilience in terms of predictive flood risk mapping at the road level. In particular, such models could help impacted communities and emergency management agencies develop better preparedness and response strategies with improved situational awareness of road inundation likelihood as an extreme weather event unfolds.

Machine Learning and Hyperspectral Imaging-Aided Forecast for the Share of Biogenic and Fossil Carbon in Solid Waste

Traditional methods for analyzing the biogenic and fossil carbon shares in solid waste are time-consuming and labor-intensive. A novel approach was developed to directly classify the carbon group and predict carbon content using the hyperspectral imaging (HSI) spectra of solid waste in conjunction with state-of-the-art tree-based machine learning models, including random forest (RF), extreme gradient boost, and light gradient boost machine (LGBM). All of the classifiers and regressors were able to achieve an accuracy above 0.95 and an R2 of 0.96 in the test set, respectively. In addition, two model interpretation approaches, the Shapley additive explanation and model explainer, were applied. The results showed that the predictions of the developed models were based on a reasonable understanding of the overtone and shake of the functional groups (C–H, N–H, and O–H). Furthermore, the developed models were validated by an external test set, which did not overlap with the data used for model construction. The RF and LGBM showed robust performance with a 0.790 accuracy for carbon group classification and a 0.806 R2 for carbon content prediction. Overall, the optimal models provided a rapid method for characterizing the biogenic carbon share in solid waste based on raw HSI spectra without preprocessing.

Market prediction using machine learning based on social media specific features

In recent years, unspecified messages posted on social media have significantly affected the price fluctuations of online-traded products, such as stocks and virtual currencies. In this study, we investigate whether information on Twitter and natural language expressions in tweets can be used as features for predicting market information, such as price changes in virtual currencies and sudden price changes. Our method is based on features created using Sentence-BERT for tweet data. These features were used to train the light-gradient boosting machine (LightGBM), a variant of the gradient boosting ensemble framework that uses tree-based machine learning models, with the target variable being a sudden change in closing price (sudden drop, sudden rise, or no sudden change). We set up a classification task with three labels using the features created by the proposed method for prediction. We compared the prediction results with and without these new features and discussed the advantages of linguistic features for predicting changes in cryptocurrency trends.

Tree-based machine learning model for visualizing complex relationships between biochar properties and anaerobic digestion

The ideal conditions for anaerobic digestion experiments with biochar addition are challenging to thoroughly study due to different experimental purposes. Therefore, three tree-based machine learning models were developed to depict the intricate connection between biochar properties and anaerobic digestion. For the methane yield and maximum methane production rate, the gradient boosting decision tree produced R2 values of 0.84 and 0.69, respectively. According to feature analysis, digestion time and particle size had a substantial impact on the methane yield and production rate, respectively. When particle sizes were in the range of 0.3–0.5 mm and the specific surface area was approximately 290 m2/g, corresponding to a range of O content (>31%) and biochar addition (>20 g/L), the maximum promotion of methane yield and maximum methane production rate were attained. Therefore, this study presents new insights into the effects of biochar on anaerobic digestion through tree-based machine learning.

Snow avalanche susceptibility mapping from tree-based machine learning approaches in ungauged or poorly-gauged regions

Mapping avalanche susceptibility is essential for disaster management, especially in ungauged or poorly-gauged regions. Most existing machine-learning methods have a shortfall in multi-source, heterogeneous spatial data requirements and overfitting. Faced with these deficiencies, this study has developed four novel tree-based machine-learning models (Catboost, LightGBM, RF, and XGBoost) to predict avalanche susceptibility, along with the data support of remote sensing images (e.g., Superview-1, Sentinel-2, and Calibrated Enhanced-Resolution Passive Microwave Daily Brightness Temperature), Meteorological data (e.g., WRF simulation), topography, and limited observation in Tianshan Mountains, China. There are 21 causative factors in the selection, training, and validation processes. Then, the SHAP value is introduced to quantify the variable-based share of contribution to the possibility of avalanches on both global and local scales. The results demonstrate the following: (1) All four models are potent in assessing avalanche susceptibility due to similar patterns, with Catboost being the best with its eight indices: Accuracy (0.9249), TPR (0.9920), FAR (0.0080), TNR (0.8904), PPR (0.8229), NPR (0.0046), CSI (0.8175), and HSS (0.8404) superior to others; (2) The Catboost provides a reliable susceptibility map illustrating moderate and high susceptible areas up to 54.12% of the total, mainly concentrated in the west and southeast; (3) Topographic factors (distance to rivers, aspect, and relative slope position) and meteorology (precipitation) are the most effective for susceptibility modeling. The results above indicate the significant potential of tree-based machine-learning with multi-source and heterogeneous spatial data in obtaining high-quality susceptibility maps without too much field observation. It is of great importance for avalanche prevention in regions with complex terrain conditions and sparse data.

Optimization of Tree-Based Machine Learning Models to Predict the Length of Hospital Stay Using Genetic Algorithm.

The length of hospital stay (LOS) is a significant indicator of the quality of patient care, hospital efficiency, and operational resilience. Considering the importance of LOS in hospital resource management, this research aims to improve the accuracy of LOS prediction using hyperparameter optimization (HPO). Expert physicians and related studies were reviewed to determine the variables affecting LOS. The electronic medical records of 200 patients in the department of internal medicine of a hospital in Iran were collected randomly. As the performance of machine learning (ML) models can vary based on the characteristics of the features, several models were applied and evaluated in this study. In particular, k-nearest neighbors (KNN), multivariate regression, decision tree (DT), random forest (RF), artificial neural network (ANN), and XGBoost have been evaluated and improved. The genetic algorithm (GA) was applied to optimize the tree-based models. In addition, the dummy coding technique, sometimes called the One-Hot encoding, was used to encode categorical features to increase prediction accuracy. Compared with other algorithms, the XGBoost model optimized by GA (XGB_GA) achieved higher accuracy and better prediction performance. The mean and median of absolute errors in the test dataset for this model were 1.54 and 1.14 days, respectively. In other words, the XGB_GA model reduced the mean absolute error by 37%, which is beneficial in the reliable design of a clinical decision support system.

Machine learning in the prediction of cardiac surgery associated acutekidney injury with early postoperative biomarkers.

To establish novel prediction models for predicting acute kidney injury (AKI) after cardiac surgery based on early postoperative biomarkers. This study enrolled patients who underwent cardiac surgery in a Chinese tertiary cardiac center and consisted of a discovery cohort (n = 452, from November 2018 to June 2019) and a validation cohort (n = 326, from December 2019 to May 2020). 43 biomarkers were screened using the least absolute shrinkage and selection operator and logistic regression to construct a nomogram model. Three tree-based machine learning models were also established: eXtreme Gradient Boosting (XGBoost), random forest (RF) and deep forest (DF). Model performance was accessed using area under the receiver operating characteristic curve (AUC). AKI was defined according to the Kidney Disease Improving Global Outcomes criteria. Five biomarkers were identified as independent predictors of AKI and were included in the nomogram: soluble ST2 (sST2), N terminal pro-brain natriuretic peptide (NT-proBNP), heart-type fatty acid binding protein (H-FABP), lactic dehydrogenase (LDH), and uric acid (UA). In the validation cohort, the nomogram achieved good discrimination, with AUC of 0.834. The machine learning models also exhibited adequate discrimination, with AUC of 0.856, 0.850, and 0.836 for DF, RF, and XGBoost, respectively. Both nomogram and machine learning models had well calibrated. The AUC of sST2, NT-proBNP, H-FABP, LDH, and UA to discriminate AKI were 0.670, 0.713, 0.725, 0.704, and 0.749, respectively. In addition, all of these biomarkers were significantly correlated with AKI after adjusting clinical confounders (odds ratio and 95% confidence interval of the third vs. the first tertile: sST2, 3.55 [2.34-5.49], NT-proBNP, 5.50 [3.54-8.71], H-FABP, 6.64 [4.11-11.06], LDH, 7.47 [4.54-12.64], and UA, 8.93 [5.46-15.06]). Our study provides a series of novel predictive models and five biomarkers for enhancing the risk stratification of AKI after cardiac surgery.

Machine-Learning Approach to Predict Total Fabrication Duration of Industrial Pipe Spools

In industrial modular construction, components fabricated in the shop are assembled into modules before being transported to construction sites. This approach offers improved construction efficiency, higher quality products, and the opportunity to work in a controlled environment that offers more predictable schedules. A distinct category of modularized construction is that of piping structures where prefabricated segments of pipes, called pipe spools, are arranged and attached to a steel structure to speed up on-site assembly. Determining when a particular pipe spool is ready for delivery to the site is affected by many factors, including material availability, workforce performance and availability, shop loading conditions, and shop capacity. These factors combine to produce a high degree of uncertainty when estimating the fabrication duration. This paper investigates the improvement of delivery time prediction of pipe spools that are fabricated in a manufacturing facility. The study’s main contribution is to propose a data-driven machine-learning model that accurately predicts the total fabrication duration in a pipe spool fabrication facility. The developed model is expected to assist production managers to plan for the appropriate workforce and material delivery requirements. The proposed approach utilizes a combination of product composition and real-time loading status of the shop to develop and examine several tree-based machine-learning models. The approach was evaluated using records of 6,989 pipe spools. The overall prediction performance was improved by a reduction of 35% in the mean absolute percentage error compared to original heuristic estimates. The results also show a strong positive correlation of 0.8 between the actual and predicted values.

Inferring Trip Destination Purposes for Trip Records Collected through Smartphone Apps

Rapid developments in smartphones and Global Positioning System (GPS) technology have provided a new method for travel data collection. People’s travel trajectories can be passively collected through smartphones with built-in GPS sensors and processed to infer important attributes of travel behavior such as trip destination purposes. This paper examines the feasibility of using discrete choice models and tree-based machine learning models for trip purpose inference. Using smartphone GPS trajectories and land-use data that are open-sourced for academic research, four models are developed to classify trip destination purposes into one of thirteen categories. The models include multinomial logit, mixed logit, random forest, and gradient boosting decision tree. Because the data set is significantly unbalanced with more than 50% of the trips being home or work trips, a two-stage modeling process is applied to identify home and work trips first and then subsequently to classify trips for the remaining discretionary purposes. The results show that both discrete choice models and machine learning models can achieve more than 70% prediction accuracies for home trips and work trips, but less than 30% prediction accuracies for most of the discretionary trip purposes. The discrete choice models can accurately reproduce the market share distribution of the trip purposes, but the machine learning models fail to do so. The results imply that as purely data-driven models, machine learning models may not be the best solution for trip purpose inference because they lack the theoretical background provided by microeconomics and human psychology that are essential in explaining people’s travel and activity choices.

A Machine Learning Approach to Support Urgent Stroke Triage Using Administrative Data and Social Determinants of Health at Hospital Presentation: Retrospective Study.

The key to effective stroke management is timely diagnosis and triage. Machine learning (ML) methods developed to assist in detecting stroke have focused on interpreting detailed clinical data such as clinical notes and diagnostic imaging results. However, such information may not be readily available when patients are initially triaged, particularly in rural and underserved communities. This study aimed to develop an ML stroke prediction algorithm based on data widely available at the time of patients' hospital presentations and assess the added value of social determinants of health (SDoH) in stroke prediction. We conducted a retrospective study of the emergency department and hospitalization records from 2012 to 2014 from all the acute care hospitals in the state of Florida, merged with the SDoH data from the American Community Survey. A case-control design was adopted to construct stroke and stroke mimic cohorts. We compared the algorithm performance and feature importance measures of the ML models (ie, gradient boosting machine and random forest) with those of the logistic regression model based on 3 sets of predictors. To provide insights into the prediction and ultimately assist care providers in decision-making, we used TreeSHAP for tree-based ML models to explain the stroke prediction. Our analysis included 143,203 hospital visits of unique patients, and it was confirmed based on the principal diagnosis at discharge that 73% (n=104,662) of these patients had a stroke. The approach proposed in this study has high sensitivity and is particularly effective at reducing the misdiagnosis of dangerous stroke chameleons (false-negative rate <4%). ML classifiers consistently outperformed the benchmark logistic regression in all 3 input combinations. We found significant consistency across the models in the features that explain their performance. The most important features are age, the number of chronic conditions on admission, and primary payer (eg, Medicare or private insurance). Although both the individual- and community-level SDoH features helped improve the predictive performance of the models, the inclusion of the individual-level SDoH features led to a much larger improvement (area under the receiver operating characteristic curve increased from 0.694 to 0.823) than the inclusion of the community-level SDoH features (area under the receiver operating characteristic curve increased from 0.823 to 0.829). Using data widely available at the time of patients' hospital presentations, we developed a stroke prediction model with high sensitivity and reasonable specificity. The prediction algorithm uses variables that are routinely collected by providers and payers and might be useful in underresourced hospitals with limited availability of sensitive diagnostic tools or incomplete data-gathering capabilities.

Tree-Based Machine Learning Models with Optuna in Predicting Impedance Values for Circuit Analysis.

The transmission characteristics of the printed circuit board (PCB) ensure signal integrity and support the entire circuit system, with impedance matching being critical in the design of high-speed PCB circuits. Because the factors affecting impedance are closely related to the PCB production process, circuit designers and manufacturers must work together to adjust the target impedance to maintain signal integrity. Five machine learning models, including decision tree (DT), random forest (RF), extreme gradient boosting (XGBoost), categorical boosting (CatBoost), and light gradient boosting machine (LightGBM), were used to forecast target impedance values. Furthermore, the Optuna algorithm is used to determine forecasting model hyperparameters. This study applied tree-based machine learning techniques with Optuna to predict impedance. The results revealed that five tree-based machine learning models with Optuna can generate satisfying forecasting accuracy in terms of three measurements, including mean absolute percentage error (MAPE), root mean square error (RMSE), and coefficient of determination (R2). Meanwhile, the LightGBM model with Optuna outperformed the other models. In addition, by using Optuna to tune the parameters of machine learning models, the accuracy of impedance matching can be increased. Thus, the results of this study suggest that the tree-based machine learning techniques with Optuna are a viable and promising alternative for predicting impedance values for circuit analysis.

Building a predictive machine learning model of gentrification in Sydney

In an era of rapid urbanisation and increasing wealth, gentrification is an urban phenomenon impacting many cities around the world. The ability of policymakers and planners to better understand and address gentrification-induced displacement hinges upon proactive intervention strategies. It is in this context that we build a tree-based machine learning (ML) model to predict neighbourhood change in Sydney. Change, in this context, is proxied by the Socioeconomic Index for Advantage and Disadvantage, in addition to census and other ancillary predictors. Our models predict gentrification from 2011 to 2016 with a balanced accuracy of 74.7 %. Additionally, the use of an additive explanation tool enables individual prediction explanations and advanced feature contribution analysis. Using the ML model, we predict future gentrification in Sydney up to 2021. The predictions confirm that gentrification is expanding outwards from the city centre. A spill-over effect is predicted to the south, west and north-west of former gentrifying hotspots. The findings are expected to provide policymakers with a tool to better forecast where likely areas of gentrification will occur. This future insight can then inform suitable policy interventions and responses in planning for more equitable cities outcomes, specifically for vulnerable communities impacted by gentrification and neighbourhood change.

The Role of Technology in the Learning Process: A Decision Tree-Based Model Using Machine Learning

Machine learning approaches may establish a complex and non-linear relationship among input and response variables for the assessment of the Basic Education Development Index (IDEB) database and show indicators that may contribute to monitoring the quality of education. This paper uses extensive experimental databases from public schools, consisting of a case study in Brazil, to analyze data such as the physical and technological structure of schools and teacher profiles. The research proposes decision tree-based machine learning models for predictions of the best attributes to positively contribute to IDEB. It employs a newly developed SHapley Additive exPlanations (SHAP) approach to classify input variables, so to identify variables that impact the most the final model; a non-probabilistic sample was used, composed from three official databases of 450 schools, and 617 teachers. Results show that the number of computers per student, teachers’ service time, broadband internet access, investments in technology training for teachers, and computer labs in schools are the variables that have the greatest effect on IDEB. The model applied shows high prediction accuracy for test data (MSE = 0.2094 and R² = 0.8991). This article contributes to improving efficiency when monitoring parameters used to measure the quality of a teaching-learning process. Doi: 10.28991/ESJ-2022-SIED-020 Full Text: PDF