New scripts to cluster based on C->S/S->C direction filter; improving the plot's...

[pingpong.git] / python_ml / plotting-dbscan-diff-metric.py

diff --git a/python_ml/plotting-dbscan-diff-metric.py b/python_ml/plotting-dbscan-diff-metric.py
new file mode 100644 (file)
index 0000000..947705d
--- /dev/null
+++ b/python_ml/plotting-dbscan-diff-metric.py
@@ -0,0 +1,132 @@
+from sklearn.cluster import DBSCAN
+from sklearn import metrics
+import matplotlib.cm as cm
+import numpy as np
+import matplotlib.pyplot as plt
+
+# metric function for clustering
+def metric(x, y):
+       # Compare 2 datapoints in array element 2 and 3 that contains C or S
+       if x[2] != y[2] or x[3] != y[3]:
+               # We are not going to cluster these together since they have different directions
+               return sys.maxsize;
+       else:
+               # Compute Euclidian distance here
+               return math.sqrt((x[0] - y[0])**2 + (x[1] - y[1])**2)
+
+# Create a subplot with 1 row and 2 columns
+fig, (ax2) = plt.subplots(1, 1)
+fig.set_size_inches(20, 20)
+
+# Read from file
+# TODO: Just change the following path and filename 
+#      when needed to read from a different file
+path = "/scratch/July-2018/Pairs3/"
+# TODO: Change the order of the files below to generate
+#              the diff plot reversedly
+device1 = "kwikset-off-phone-side"
+device2 = "kwikset-on-phone-side"
+filename1 = device1 + ".txt"
+filename2 = device2 + ".txt"
+plt.ylim(0, 2000)
+plt.xlim(0, 2000)
+
+# Number of triggers
+trig = 50
+
+# PLOTTING FOR DEVICE ON EVENT
+# Read and create an array of pairs
+with open(path + filename1, "r") as pairs:
+       pairsArr1 = list()
+       pairsSrcLabels1 = list()
+       for line in pairs:
+               # We will see a pair and we need to split it into xpoint and ypoint
+               xpoint, ypoint, srcHost1, srcHost2, src1, src2 = line.split(", ")
+               # Assign 1000 for client and 0 for server to create distance
+               src1Val = 1000 if src1 == 'C' else 0
+               src2Val = 1000 if src2 == 'C' else 0
+               pair = [int(xpoint), int(ypoint), int(src1Val), int(src2Val)]
+               pairSrc = [int(xpoint), int(ypoint), srcHost1, srcHost2, src1, src2]
+               # Array of actual points
+               pairsArr1.append(pair)
+               # Array of source labels
+               pairsSrcLabels1.append(pairSrc)
+
+# PLOTTING FOR DEVICE ON EVENT
+# Read and create an array of pairs
+with open(path + filename2, "r") as pairs:
+       pairsArr2 = list()
+       pairsSrcLabels2 = list()
+       for line in pairs:
+               # We will see a pair and we need to split it into xpoint and ypoint
+               xpoint, ypoint, srcHost1, srcHost2, src1, src2 = line.split(", ")
+               # Assign 1000 for client and 0 for server to create distance
+               src1Val = 1000 if src1 == 'C' else 0
+               src2Val = 1000 if src2 == 'C' else 0
+               pair = [int(xpoint), int(ypoint), int(src1Val), int(src2Val)]
+               pairSrc = [int(xpoint), int(ypoint), srcHost1, srcHost2, src1, src2]
+               # Array of actual points
+               pairsArr2.append(pair)
+               # Array of source labels
+               pairsSrcLabels2.append(pairSrc)         
+
+diff12 = [i for i in pairsArr1 if i not in pairsArr2]
+diff12SrcLabels = [i for i in pairsSrcLabels1 if i not in pairsSrcLabels2]
+
+X = np.array(diff12);
+
+# Compute DBSCAN
+# eps = distances
+# min_samples = minimum number of members of a cluster
+db = DBSCAN(eps=10, min_samples=trig - 45).fit(X)
+core_samples_mask = np.zeros_like(db.labels_, dtype=bool)
+core_samples_mask[db.core_sample_indices_] = True
+labels = db.labels_
+
+# Number of clusters in labels, ignoring noise if present.
+n_clusters_ = len(set(labels)) - (1 if -1 in labels else 0)
+
+# Black removed and is used for noise instead.
+unique_labels = set(labels)
+
+colors = [plt.cm.Spectral(each)
+             for each in np.linspace(0, 1, len(unique_labels))]
+for k, col in zip(unique_labels, colors):
+       cluster_col = [1, 0, 0, 1]
+       if k == -1:
+           # Black used for noise.
+           col = [0, 0, 0, 1]
+
+       class_member_mask = (labels == k)
+
+       # print("Unique label: " + str(k) + " with freq: " + str(labels.tolist().count(k)))
+       xy = X[class_member_mask & core_samples_mask]
+       plt.plot(xy[:, 0], xy[:, 1], 'o', markerfacecolor=tuple(cluster_col),
+                markeredgecolor='k', markersize=10)
+
+       xy = X[class_member_mask & ~core_samples_mask]
+       plt.plot(xy[:, 0], xy[:, 1], 'o', markerfacecolor=tuple(col),
+                markeredgecolor='k', markersize=6)
+
+# Print lengths
+count = 0
+for pair in diff12:
+       if labels[count] == -1:
+               plt.text(pair[0], pair[1], str(pair[0]) + ", " + str(pair[1]), fontsize=10)
+       else:
+       # Only print the frequency when this is a real cluster
+               plt.text(pair[0], pair[1], str(pair[0]) + ", " + str(pair[1]) + 
+                       " - Freq:" + str(labels.tolist().count(labels[count])), fontsize=10)
+       count = count + 1
+
+# Print source-destination labels
+count = 0
+for pair in diff12SrcLabels:
+       # Only print the frequency when this is a real cluster
+       plt.text(pair[0], pair[1], str(pair[4]) + "->" + str(pair[5]))
+       count = count + 1
+
+plt.title(device1 + ' - diff - ' + device2)
+plt.show()
+
+