#randomforest — Public Fediverse posts

Comparing DESIS Hyperspectral and Landsat 10 Simulated Superspectral Data for Crop Type Classification in California’s Central Valley
--
https://doi.org/10.3390/rs18142282 <-- shared paper
--
https://www.usgs.gov/publications/comparing-desis-hyperspectral-and-landsat-10-simulated-superspectral-data-crop-type <-- shared USGs publication page
--
H/T @USGS
“How can we get better at classifying crops from space? 🛰️🌽
Some types of satellite imagery are better at distinguishing crops than others. A USGS study compared two approaches and found one produced more accurate crop maps.
Here's what the researchers found:
• Hyperspectral imagery achieved 86% accuracy, compared to 75% for simulated superspectral imagery.
• Using just 14 carefully selected hyperspectral bands produced nearly the same results as using all 60 DLR Earth Sensing Imaging Spectrometer (DESIS) bands, showing those wavelengths contain much of the information needed to distinguish crop types.
• Using superspectral imagery on the upcoming Landsat 10 mission will allow for routine tracking of crops and enhance finer crop mapping.
• The findings help scientists identify which wavelengths provide the most useful information for crop mapping and future remote sensing applications.
Better crop maps will help governments and scientists track global agriculture, monitor the current crop season, and study agricultural trends…”
--
“HIGHLIGHTS:
• What are the main findings?
- The 14 DESIS hyperspectral narrowbands (10 nm) aligned with the Landsat 10 (formerly Landsat Next) spectral dataset produced similar accuracy results to the full 60-band DESIS hyperspectral dataset for classifying crop types. These 14 DESIS narrowbands resulted in higher accuracy than the 14 simulated Landsat 10 superspectral broadbands.
- When using DESIS narrowbands, Support Vector Machine (SVM) resulted in higher accuracy than Random Forest (RF).
• What are the implications of the main findings?
- A carefully selected set of 14 DESIS hyperspectral narrowbands (10 nm) can achieve classification accuracy comparable to those obtained using all 60 DESIS narrowbands across the 400–1000 nm range. These 14 strategically positioned narrowbands classified crop types with higher classification accuracy than the corresponding 14 Landsat 10 superspectral broadbands within the same spectral range.
- This study underscores the importance of multi-temporal imagery across the full crop-growing season for achieving more detailed and accurate crop type classifications. Such temporal coverage is more feasible with the planned Landsat 10 routine acquisition of broadband imagery than with task-based hyperspectral collections…”
#hyperspectral #superspectral #optimalbands #randomforest #supportvectormachine #agriculture #crops #croptype #classifaction #croplands #California #CentralValley #GIS #spatial #mapping #remotesensing #earthobservation #imagery #DESIS #Landsat #Landsat10 #satellite #spatialanalysis #spatiotemporal #global #AI #machinelearning #model #modeling #SupportVectorMachine #SVM #RandomForest #RF #GoogleEarthEngine
@USGS

J'ai finalement implémenté une approche par apprentissage (random forest) pour prévoir (on dit inférer) l'évolution de la température des capteurs de température intérieur et extérieur à partir des données de prévision météo locales et de la position du soleil.
L'idée est de pouvoir estimer l'heure à laquelle on va pouvoir ouvrir les fenêtres. J'ai déjà une alarme qui me réveille quand c'est le cas, mais je suis content d'avoir une vague idée en me couchant de l'heure de ce réveil.

Évidemment, avec 10 jours de données, pour l'instant le modèle est un peu spécialisé en début de canicules, mais plus on va avancer, plus j'aurai de données collectées pour améliorer le modèle.

C'est bien sûr open-source :
https://codeberg.org/jmfavreau/micro-climat-predict

#ChroniqueClimatique #canicule #simulation #RandomForest #zigbee #HomeAssistant

Mapping Deforestation Probability And Understanding The Forest Dynamics In Gazipur, Bangladesh
--
https://doi.org/10.1016/j.envc.2026.101568 <-- shared paper
--
"ABSTRACT: Deforestation is a spiralling environmental catastrophe with impervious results for biodiversity, climate change, and human livelihoods, specifically in tropical regions. Being a tropical country, Bangladesh has experienced approximately 40% loss of its forest cover, at Gazipur since 1930, which contains about 86% of the country's Sal (Shorea robusta) forest, ranging approximately 4,300 hectares per year (2001–2010) to over 19,500 hectares per year (2011–2020), exemplifying an intensification of nearly 353%. The objective of this study is to map deforestation probability at the Gazipur district of Dhaka Division, Bangladesh, by utilising machine learning algorithms along with multi-source geospatial data, with the purpose of identifying high-risk zones and facilitating evidence-based forest governance, land-use development, and prioritizing conservation areas. This study integrated twelve conditioning factors, including biophysical, landscape, and anthropogenic. To identify susceptible zones the study trained and assessed five machine learning algorithms; RF, XGBoost, ANN, NB, and MLP and validating the result through different metrics like sensitivity, specificity, precision, accuracy, F1-score, AUC. The performance of the models was evaluated using Wilcoxon signed-rank tests and marginal response curves (MRC) were used to understand factor contributions. In the result, RF achieved highest performance with accuracy of 84% and AUC of 0.93, followed by XGBoost at 83% accuracy and 0.92 AUC. Rainfall and population density were most dominant conditioning factors among models. Pairwise statistical testing resulted that ensemble-based algorithms (RF, XGBoost) generated statistically comparable and significantly higher predictions compared to NB and MLP. Spatial probability maps indicate areas of high and very high risk in the south-western and north-eastern upazilas. The results can be applicable for forest management authorities, urban planners, and policymakers, and correspond with SDG Indicator 15. An inclusive governance framework containing land zoning, ecological area identification, and compliance with industrial EIA is proposed to persuade probability maps into adaptive forest management strategies…”
#deforestation #probability #machinelearning #algorithms #AI #Gazipur #Bangladesh #GIS #spatial #mapping #spatialanalysis #spatiotemporal #rainfall #precipitation #humanimpacts #populationpressure #risk #prediction #RandomForest #conservation #restoration #environment #biodiversity, #climatechange #human #livelihood #tropical #forestcover #sal #forest #vegetation #tree #upazila #spatialprobability #geostatistics #forestmanagement #planning #policy #urbanplanners #governance #zoning #ecology #habitat

Comparative Analysis Of Seagrass Biophysical Properties Mapping Using Multi-Resolution Satellite Imagery And Machine Learning In The Shallow Waters Of Teluk Pandan, Lampung, Indonesia
--
https://doi.org/10.1016/j.rsase.2026.102002 <-- shared paper
--
#GIS #spatial #mapping #Seagrass #monitoring #conservation #accuracy #vegetation #biophysical #benthic #habitat #composition #carbonmapping #Randomforest #ExtremeGradientBoosting #XGBoost algorithms #Sentinel2 #remotesensing #sensor #shallowwater #sentinel #PlanetScope #satellite #TelukPandan #Lampung #Indonesia #AI #deeplearning #machinelearning #model #modeling #water #marine #ocean #habitat #ecosystem #spatialanalysis

【🎉Latest accepted article】
Enhancing Forest Biomass Estimation with Synthetic #AirborneLaserScanning via Voxel-based Forest Reconstruction

#AbovegroundBiomass | #LiDARSimulation | #VirtualForest | #RandomForest

https://doi.org/10.1093/jpe/rtag051

The Statistics Globe Hub starts in 3 days, and I would like to give you a short preview of the first module "Feature Selection Using Random Forest."

Interested in joining the Hub? You can find more information here: https://statisticsglobe.com/hub

#statistics #datascience #ai #rstats #python #randomforest

Geomorphic Factors Impact Groundwater Levels More Than Harvesting in a Coast Redwood Forest
--
https://doi.org/10.1002/hyp.70302 <-- shared paper
--
#water #hydrology #groundwater #stream #baseflow #plant #gegetation #transpiration #growingseason #forest #management #watershed #mediterraneanclimate #California #redwood #harvesting #soils #geomorphology #aquiferdepth #slope #randomforest #model #modeling #coast #coastal #rainfill #precipitation #infiltration
#USFS

🌳 Random Forests and Living Trees

English translation of my earlier article on applying satellite imagery and machine learning to map urban land cover.

What started as a local research project in Kryvyi Rih turned into something much larger — the results sparked a heated discussion among residents, officials, and industry representatives about the real condition of green buffers around large industrial sites.

The methodology developed during that work is still being used today — adapted for new environmental and urban projects.

🔗 https://www.datastory.org.ua/random-forests-and-living-trees/

#RemoteSensing #MachineLearning #LandCoverMapping #UrbanEcology #EnvironmentalMonitoring #RandomForest #GeospatialAnalysis #GIS #RStats #SAGAGIS #QGIS #IndependentResearch #OpenSource #EnvironmentalDataScience #KryvyiRih #LULC

Assessment Of Snow Cover Dynamics And The Effects Of Environmental Drivers In High Mountain Ecosystems
--
https://doi.org/10.1016/j.eiar.2025.107969 <-- shared paper
--
#GIS #spatial #mapping #remotesensing #earthobservation #snow #ice #snowcover #dynamics #climatechange #mountains #ecosystems #spatialanalysis #spatiotemporal #MODIS #model #modeling #extremeweather #water #hydrology #climate #zones #trendanalysis #linearregression #RandomForest #cryosphere

Avalanche Debris Detection From Sentinel-2 Data Using Fuzzy Machine Learning And Colour Spaces For The Indian Himalaya
--
https://doi.org/10.1080/2150704X.2025.2488532 <-- shared paper
--
#GIS #spatial #mapping #snowavalanches #snow #avalanches #machinelearning #fuzzyclassification #SVM #AI #randomforest #model #modeling #forecasting #risk #hazard #massmovement #engineeringgeology #remotesensing #earthobservation #imagery #spatialanalysis #spatiotemporal #change #debris #detection #satellite #sentinel #Himalaya #Himalayas #performance

This series of videos on machine learning algorithms (Lab a through Lab d, so far) by Courage Kamusoko are the best explanations I've seen yet. How the models actually work, their strengths and weaknesses, what you are actually solving for when you tune the hyperparameters, and examples in Python. https://www.youtube.com/@couragekamusoko5689/videos
#SVM #KNN #DecisionTree #RandomForest

If I were running a blog on applying #randomforest models to various problems, I would call it The Statistical Lumberjack 🤔

I scaled up the popular Palmer Penguins machine learning dataset from 344 rows to 100k rows using adversarial random forest, with an accuracy of 88%.

Now, you have more rows of data with which to train your classification models.

You can download it here, along with R & Python scripts, to load and view the dataset: https://ieee-dataport.org/documents/palmer-penguins-100k-0

Have a dataset you want to scale up? Say hello!

#machinelearning #randomforest #rstats #python #datascience #datasets #syntheticdatageneration #ai

Essentially, each path in a #RandomForest to a leaf indicates that a number of training examples satisfy a sequence of constraints (from the splits). Inferring training data boils down to finding a set of examples satisfying all these constraints, a bit like placing numbers on a Sudoku...

"Why do Random Forests Work? Understanding Tree Ensembles as Self-Regularizing Adaptive Smoothers"

https://arxiv.org/abs/2402.01502

'... Despite their remarkable effectiveness and broad application, the drivers of success underlying ensembles of trees are still not fully understood. In this paper, we highlight how interpreting tree ensembles as adaptive and self-regularizing smoothers can provide new intuition and deeper insight to this topic...'

#MachineaLearning #ML #RandomForest

One question for the #MachineLearning people: what approach do you use to determine if a decision trees or a random forest approach should work better? Do you simply try both approaches and use whatever seems to work better?

According to what I read, decision trees are more prone to overfitting, while random forest is a more complex approach. Which means little to me 😅

#ml #DecisionTrees #RandomForest #Overfitting

Here is an example of using #RandomForests 🌳🌳 for #PixelClassification 🖼️ in #Python 🐍, using @napari for labeling ✍️

🌎 https://www.fabriziomusacchio.com/blog/2023-06-23-_random_forests_pixel_classifier/

#RandomForest #Napari #MachineLearning #ImageProcessing #Bioimage #BioimageAnalysis

Ever wondered how #DecisionTrees and #RandomForests 🌳🌳 are related? Here is a quick #tutorial that compares both methods in terms of #classification and #regression ✌️

🌎 https://www.fabriziomusacchio.com/blog/2023-06-22-_decision_trees_vs_random_forests/

#DecisionTree #RandomForest #MachineLearning

How to use #randomForest for predicting community interactions

https://onlinelibrary.wiley.com/doi/10.1111/ecog.06619

New paper from @GlobEcoFlinders

'Predicting predator–prey interactions in terrestrial endotherms using #randomForest'

https://onlinelibrary.wiley.com/doi/10.1111/ecog.06619

Area Estimation of Mango and Coconut Crops using Machine Learning in Hesaraghatta Hobli of Bengaluru Urban District, Karnataka [India]
--
https://doi.org/10.58825/jog.2023.17.1.75 <-- shared paper
--
#GIS #spatial #mapping #spatialanalysis #Mango #Coconut #Areaestimation #Machinelearning #ConvolutionalNeuralNetwork #CNN #RandomForest #RF #agricultural #agriculture #farming #estimates #Karnataka #India #acreage #production #horticulture #commodities #export #remotesensing #satellite #GoogleEarthEngine #GEE #Copernicus #Sentinel2 #algorithms #learning #google #ml

Our paper about the comparison of #machineLearning and regression analysis in predicting #sexual reoffenses will be published in "#Assessment"

Guess what? "#RandomForest Does Not Outperform Logistic Regression in the Prediction of Sexual Recidivism"

With Sonja Etzler, @[email protected], @florianpargent

Preprint: https://psyarxiv.com/z6ky2

#ML #preprint #psychology

Interesting new study estimating the #replicability of published research in #psychology over the past 20 years.

The paper includes *nearly all papers* published in six top psychology #journals over last 2 decades.

https://www.pnas.org/doi/10.1073/pnas.2208863120

The researchers used a #MachineLearning model (#RandomForest & logistic regression ensemble) to estimate the replication likelihood of over 14,000 #articles from 2000-2019 in six subfields of psychology.

#Science #Academia #Research #Fediverse #OpenAccess