Testing AutoEps

df-aggregator.py

@@ -37,6 +37,7 @@ from multiprocessing import Process, Queue
 from bottle import route, run, request, get, post, put, response, redirect, template, static_file
 from bottle.ext.websocket import GeventWebSocketServer, websocket
 
+
 DBSCAN_Q = Queue()
 DBSCAN_WAIT_Q = Queue()
 DATABASE_EDIT_Q = Queue()
@@ -227,6 +228,32 @@ def do_dbscan(X, epsilon, minsamp):
     if not DBSCAN_WAIT_Q.empty():
         DBSCAN_WAIT_Q.get()
 
+####################################
+# Autocalculate the best eps value.
+####################################
+def autoeps_calc(X):
+    X = X[:min(2000, len(X)):2]
+    min_distances = []
+    for x in X:
+        distances = []
+        for y in X:
+            distance = math.sqrt(sum([(a - b) ** 2 for a, b in zip(x, y)]))
+            if distance > 0:
+                distances.append(distance)
+        min_distances.extend(np.sort(distances)[0:3].tolist())
+
+    sorted_distances = np.sort(min_distances).tolist()
+    try:
+        for x1, y1 in enumerate(sorted_distances):
+            x2 = x1 + 1
+            y2 = sorted_distances[x2]
+            m = (y2 - y1) / (x2 - x1)
+            if m > 0.005:
+                # print(f"Slope: {round(m, 3)}, eps: {y1}")
+                return y1
+    except IndexError:
+        return 0
+
 ###############################################
 # Computes DBSCAN Alorithm is applicable,
 # finds the mean of a cluster of intersections.
@@ -254,6 +281,9 @@ def process_data(database_name, epsilon, min_samp):
             if epsilon > 0:
                 X = StandardScaler().fit_transform(intersect_array[:,0:2])
                 n_points = len(X)
+                autoeps = autoeps_calc(X)
+                min_samp = max(3, round(0.05 * n_points, 0))
+                print(f"min_samp: {min_samp}, eps: {autoeps}")
                 size_x = sys.getsizeof(X)/1024
                 print(f"The dataset is {size_x} kilobytes")
                 print(f"Computing Clusters from {n_points} intersections.")
@@ -261,7 +291,8 @@ def process_data(database_name, epsilon, min_samp):
                     print("Waiting for my turn...")
                     time.sleep(1)
                 starttime = time.time()
-                db = Process(target=do_dbscan,args=(X,epsilon,min_samp))
+                db = Process(target=do_dbscan,args=(X,autoeps,min_samp))
+                # db = Process(target=do_dbscan,args=(X,epsilon,min_samp))
                 db.daemon = True
                 db.start()
                 try:

views/cesium.tpl

@@ -247,13 +247,21 @@
       const epsslider = document.getElementById("epsilonRange");
       const minpointslider = document.getElementById("minpointRange");
       const intersect_en = document.getElementById("intersect_en");
+      const clustering_en = document.getElementById("clustering_en");
 
       if(minpointslider !== null) {
         parameter += "minpts="+minpointslider.value+"&";
       }
-      if(epsslider !== null) {
-        parameter += "eps="+epsslider.value+"&";
+      if (clustering_en !== null) {
+        if (clustering_en.checked) {
+          parameter += "eps=1&";
+        } else {
+          parameter += "eps=0&";
+        }
       }
+      // if(epsslider !== null) {
+      //   parameter += "eps="+epsslider.value+"&";
+      // }
       if (intersect_en !== null) {
         if (intersect_en.checked) {
           parameter += "plotpts=true"+"&";
@@ -392,7 +400,7 @@
       </span>
       <span class="slidevalue" id="confidence"></span>
     </div>
-    <div class="tooltip">
+    <!-- <div class="tooltip">
       <span class="tooltiptext">Epsilon:<br>
         Maximum distance between neighboring points in a cluster. Set to 0 to disable clustering.<br>
         Disabling clustering will plot all intersections and may cause longer load times.</span>
@@ -407,6 +415,17 @@
         <input name="minpointValue" type="range" min="0" max="300" step="5" value="{{minpoints}}" class="slider" id="minpointRange">
       </span>
       <span class="slidevalue" id="minpoints"></span>
+    </div> -->
+    <div style="width: 600px">
+      <span class="tooltip">
+        <span class="slidetitle"><h4>Clustering:</h4></span>
+        <span class="slidespan" style="text-align:left; width: 100px;margin: 5px;">
+        <label class="switch">
+          <input id="clustering_en" name="clustering_en" {{"checked" if epsilon > 0 else ""}} type="checkbox" onchange="updateParams()">
+          <span class="switchslider round"></span>
+        </label></span>
+        <span class="tooltiptext">Turns clustering on or off. Clustering On will draw ellipses.</span>
+      </span>
     </div>
     <div style="width: 600px">
       <span class="tooltip">
@@ -433,25 +452,26 @@
     var confoutput = document.getElementById("confidence");
     confoutput.innerHTML = confslider.value;
 
-    var epsslider = document.getElementById("epsilonRange");
-    var epsoutput = document.getElementById("epsilon");
-    epsoutput.innerHTML = epsslider.value;
+    // var epsslider = document.getElementById("epsilonRange");
+    // var epsoutput = document.getElementById("epsilon");
+    // epsoutput.innerHTML = epsslider.value;
 
-    var minpointslider = document.getElementById("minpointRange");
-    var minpointoutput = document.getElementById("minpoints");
-    minpointoutput.innerHTML = minpointslider.value;
+    // var minpointslider = document.getElementById("minpointRange");
+    // var minpointoutput = document.getElementById("minpoints");
+    // minpointoutput.innerHTML = minpointslider.value;
 
     var rx_enable = document.getElementById("rx_en");
 
     var intersect_en = document.getElementById("intersect_en");
+    // var clustering_en = document.getElementById("clustering_en");
 
     // Update the current slider value (each time you drag the slider handle)
-    epsslider.oninput = function() {
-      epsoutput.innerHTML = this.value;
-    }
-    epsslider.onpointerup = function() {
-      updateParams("");
-    }
+    // epsslider.oninput = function() {
+    //   epsoutput.innerHTML = this.value;
+    // }
+    // epsslider.onpointerup = function() {
+    //   updateParams("");
+    // }
     powerslider.oninput = function() {
       poweroutput.innerHTML = this.value;
     }
@@ -464,12 +484,12 @@
     confslider.onpointerup = function() {
       updateParams("minconf="+this.value);
     }
-    minpointslider.oninput = function() {
-      minpointoutput.innerHTML = this.value;
-    }
-    minpointslider.onpointerup = function() {
-      updateParams("");
-    }
+    // minpointslider.oninput = function() {
+    //   minpointoutput.innerHTML = this.value;
+    // }
+    // minpointslider.onpointerup = function() {
+    //   updateParams("");
+    // }
 
     rx_enable.onchange = function() {
       if (rx_enable.checked) {
@@ -481,11 +501,6 @@
 
     intersect_en.onchange = function() {
       updateParams("");
-      // if (intersect_en.checked) {
-      //   updateParams("plotpts=true");
-      // } else {
-      //   updateParams("plotpts=false");
-      // }
     }
 
   </script>