Adding in training of dual model recognition system.

train_wand_model.py

@@ -0,0 +1,76 @@
+from glob import glob
+import numpy as np
+import scipy as sp
+import pandas as pd
+
+
+
+from sklearn.cluster import KMeans
+from sklearn.metrics import accuracy_score, confusion_matrix
+
+import os
+
+from sklearn.ensemble import RandomForestClassifier
+
+og_data_columns = 4
+
+
+def apply_k_means_clustering(data, kmeans):
+    hist = np.zeros((len(kmeans.cluster_centers_),))
+    for x in kmeans.predict(data):
+        hist[x]+=1
+    print(hist)
+    return hist
+
+def quantize_flattened_data(data, length):
+    # print("quantizing the data")
+    # print(data)
+    quantile_length = length*og_data_columns
+    rows = int(len(data)/quantile_length)
+    return np.resize(data, (rows, quantile_length))
+
+def load_data(data_dir):
+    data = []
+    for root, dirs, files in os.walk(data_dir):
+        if root is data_dir:
+            continue
+        for file in files:
+            data.append([os.path.basename(root), pd.read_csv(os.path.join(root, file)).values.flatten()])
+    return pd.DataFrame(data, columns=["exercise", "flat_data"])
+
+train_data = load_data("./train")
+test_data = load_data("./test")
+# val_data = load_data("./val")
+
+quant_len = 3
+
+train_data["quantized_data"] = train_data["flat_data"].apply(lambda x : quantize_flattened_data(x, quant_len))
+test_data["quantized_data"] = test_data["flat_data"].apply(lambda x : quantize_flattened_data(x, quant_len))
+
+
+linked_data = train_data["quantized_data"][0]
+
+for x in train_data["quantized_data"]:
+    linked_data = np.append(linked_data, x, axis=0)
+
+print(linked_data)
+
+k_means_model = KMeans(n_clusters=25, n_init=15, max_iter= 500).fit(linked_data)
+
+print(k_means_model.cluster_centers_)
+
+train_data["histogram"] = train_data["quantized_data"].apply(lambda x: apply_k_means_clustering(x, k_means_model))
+test_data["histogram"] = test_data["quantized_data"].apply(lambda x: apply_k_means_clustering(x, k_means_model))
+print(train_data)
+
+clf = RandomForestClassifier(n_estimators=50)
+print(train_data["histogram"].values)
+
+clf.fit([x for x in train_data["histogram"]], train_data["exercise"].values)
+results = clf.predict([x for x in test_data["histogram"]])
+print(results)
+print(test_data["exercise"].values)
+
+print(accuracy_score( test_data["exercise"].values,results ))
+print(confusion_matrix( test_data["exercise"].values,results))
+