Random Forest Classification

/*set up function to mask clouds using the Sentinel-2 QA band*/
function maskS2clouds(image) {
 
  // select QA band for Sentinel-2
  var qa = image.select('QA60');
 
  // Bits 10 and 11 are clouds and cirrus, respectively.
  var cloudBitMask = 1 << 10;
  var cirrusBitMask = 1 << 11;
 
  // Both flags should be set to zero, indicating clear conditions.
  var mask = qa.bitwiseAnd(cloudBitMask).eq(0)
  	.and(qa.bitwiseAnd(cirrusBitMask).eq(0));
  	
  // Mask the cloudy image and scale the unit by dividing 10000
  return image.updateMask(mask).divide(10000);
}
 
//lat & lon for RTP
var point = ee.Geometry.Point(-79.04, 35.91);
 
// Load Sentinel-2 surface reflectance data and
// filter the images with time, place and cloud conditions
var dataset = ee.ImageCollection('COPERNICUS/S2_SR')
              	.filterDate('2020-05-01', '2020-10-30')
              	.filterBounds(point) //filter by location
              	// Pre-filter to get less cloudy granules.
                  .filter(ee.Filter.lt('CLOUDY_PIXEL_PERCENTAGE', 10))
              	.map(maskS2clouds);
         	
var image = dataset.first();
var rgbVis = {
  min: 0.0,
  max: 0.2,
  bands: ['B4', 'B3', 'B2'],
};
 
Map.centerObject(image,10);
//load both the true color and the color infrared
Map.addLayer(image, rgbVis, 'Sentinel-2 RGB');
Map.addLayer(image, {bands:['B8', 'B4', 'B3'], min:0, max:1}, 'Sentinel-2 CIR');

// calculate NDVI and add it to image
var ndvi = image.normalizedDifference(['B8', 'B4']).rename('NDVI');
image = image.addBands(ndvi)
//display the NDVI
Map.addLayer(ndvi, {min: 0.0, max: 1}, "NDVI");

// specify the feature list used in the following
var feature_bands = ['B2','B3','B4','B8','NDVI'];

// Classification training samples were obtained by hand-drawn geometry  
// Merge samples together
var feature_samples = Forest.merge(Grass).merge(Builtup).merge(Barren).merge(Water);

// add a column of random numbers [0-1]
feature_samples = feature_samples.randomColumn('random',1);
 
// Randomly split the data into 70% for training, and 30% for testing
 
var training_data =feature_samples.filter(ee.Filter.lte('random', 0.70));
var testing_data =feature_samples.filter(ee.Filter.gt('random', 0.70));

// Sample the pixels of the input imagery to get a FeatureCollection of training data.
var training = image.select(feature_bands).sampleRegions({
  collection: training_data, //
  properties: ['landcover'], // The list of properties to copy from each input feature
  scale: 10  // pixel scale
});
 
var testing = image.select(feature_bands).sampleRegions({
  collection: testing_data, //
  properties: ['landcover'], // The list of properties to copy from each input feature
  scale: 10  // pixel scale
});

// Make a Random Forest classifier and train it.
var trained_classifier = ee.Classifier.smileRandomForest(10) // default parameters for random forest and number of trees set to 10
  .train({
  features: training,
  classProperty: 'landcover',
  inputProperties: feature_bands
});

// Training accuracy evaluation
var trainAccuracy = trained_classifier.confusionMatrix();
print('Overall Accuracy for Training: ', trainAccuracy.accuracy());

// Testing accuracy evaluation
// Perform the Random Forest on the testing data
var RFtesting = testing
    .classify(trained_classifier, 'predict');
 
// calculate error matrix based on actual and predicted land cover types
// only type values in the list [1,2,3,4,5] are used
var testAccuracy = RFtesting.errorMatrix('landcover','predict',[1,2,3,4,5])
print('Overall Accuracy for Testing: ', testAccuracy.accuracy());

// Classify the input imagery.
var classified_image = image.select(feature_bands).classify(trained_classifier);

// Define a palette for the land classification.
var palette = [
   '008000', // Forest (1)
  '8df4ff', // Grass (2)
  'f56952', // Buildup (3)
  '9d7d73', // Barren (4)
  '514eff'// Water (5)
];
// plot your classification
Map.addLayer(classified_image, {min: 1, max: 5, palette: palette}, 'Classification');

// Import the NLCD collection.
var dataset = ee.ImageCollection('USGS/NLCD_RELEASES/2021_REL/NLCD');
// The collection contains images for the 2021 year release and the full suite
// of products.
print('Products:', dataset.aggregate_array('system:index'));
// Filter the collection to the 2021 product.
var nlcd2021 = dataset.filter(ee.Filter.eq('system:index', '2021')).first();
// Each product has multiple bands for describing aspects of land cover.
print('Bands:', nlcd2021.bandNames());
// Select the land cover band.
var landcover = nlcd2021.select('landcover');
// Display land cover on the map.
Map.setCenter(-79.04, 35.91, 9);
Map.addLayer(landcover, null, 'NLCD Landcover 2021');

// Add classification legend
// set position of panel
var legend = ui.Panel({
  style: {
	position: 'bottom-left',
	padding: '8px 15px'
  }
});
 
// Create legend title
var legendTitle = ui.Label({
  value: 'Legend',
  style: {
	fontWeight: 'bold',
	fontSize: '18px',
	margin: '0 0 4px 0',
	padding: '0'
	}
});
 
// Add the title to the panel
legend.add(legendTitle);
	
// Creates and styles 1 row of the legend.
var makeRow = function(color, name) {
  	
  	// Create the label that is actually the colored box.
  	var colorBox = ui.Label({
    	style: {
          backgroundColor: '#' + color,
      	// Use padding to give the box height and width.
      	padding: '8px',
      	margin: '0 0 4px 0'
    	}
  	});
  	
  	// Create the label filled with the description text.
  	var description = ui.Label({
    	value: name,
    	style: {margin: '0 0 4px 6px'}
  	});
  	
  	// return the panel
  	return ui.Panel({
    	widgets: [colorBox, description],
    	layout: ui.Panel.Layout.Flow('horizontal')
  	});
};
 
// Add color and name
 
var classType = ['Forest','Grass','Built-up','Barren','Water']
 
classType.forEach(function(ID, index) {
legend.add(makeRow(palette[index], classType[index]));
});
 
// Add legend to map
Map.add(legend);

Export.image.toDrive(classified_image, 'classificationlab');