notebook.community

Edit and run 





In [1]:



    


%matplotlib inline

import xgboost as xgb

import numpy as np
import pandas as pd

import matplotlib as mpl
import matplotlib.pyplot as plt
import seaborn as sns

import operator
import os
import pickle

CONFIGURATIONS





In [14]:



    


sns.set(context="notebook", style="ticks", font_scale=1.5)
figsize=(10,7.5)

PATHS





In [3]:



    


FEATURES_PATH = "E:\\eaglesense\\data\\topviewkinect\\all\\eval-chi2"



    











In [4]:



    


FIGS_DIR = "figs"
FIGS_HIGH_DIR = "figs/high"

if not os.path.exists(FIGS_DIR):
    os.makedirs(FIGS_DIR)
if not os.path.exists(FIGS_HIGH_DIR):
    os.makedirs(FIGS_HIGH_DIR) 
    
FIG_TEMPLATE = FIGS_DIR + "/{}.png"
FIG_HIGH_TEMPLATE = FIGS_HIGH_DIR + "/{}.pdf"

LOAD DATA





In [5]:



    


features_csv = "{path}_features.csv".format(path=FEATURES_PATH)
features_df = pd.read_csv(features_csv)



    











In [6]:



    


labels_csv = "{path}_labels.csv".format(path=FEATURES_PATH)
labels_df = pd.read_csv(labels_csv)

ACTIVITIES





In [7]:



    


ACTIVITIES = ["Standing", "Sitting", "Pointing", "Phone", "Tablet", "Paper"]

FEATURES





In [8]:



    


features_vector = features_df.drop(["subject"], axis=1)



    











In [9]:



    


output_vector = labels_df.drop(["subject"], axis=1)



    











In [10]:



    


num_features = features_vector.shape[1]
num_features



    


















    
Out[10]:








72

















In [11]:



    


features_corr = features_vector.corr()



    











In [12]:



    


# new_features_df = pd.DataFrame(columns=["feature_x", "feature_y", "correlation"])
# for feature_name, correlation_values in features_corr.iterrows():
#     for feature in features:
#         new_features_df.loc[len(new_features_df)] = [feature_name, feature, correlation_values[feature]]
# new_features_df.head()
# new_features_df_pivoted = new_features_df.pivot("feature_x", "feature_y", "correlation")
# new_features_df_pivoted.head()



    











In [13]:



    


FIG_FEATURES_CORRELATION = "features_correlation"

fig, ax = plt.subplots(figsize=figsize)
# sns.clustermap(new_features_df_pivoted)
x = sns.heatmap(features_corr, ax=ax, xticklabels=False, yticklabels=False)
sns.despine()
plt.yticks(rotation=0)

plt.savefig(FIG_TEMPLATE.format(FIG_FEATURES_CORRELATION), bbox_inches="tight")
plt.savefig(FIG_HIGH_TEMPLATE.format(FIG_FEATURES_CORRELATION), bbox_inches="tight")

OVERVIEW





In [14]:



    


X = features_vector.values



    











In [15]:



    


y = output_vector[["activity"]].values



    











In [16]:



    


y_bins = np.bincount(np.squeeze(y))



    











In [17]:



    


activities_df = pd.DataFrame({
    "activity": ACTIVITIES,
    "size": y_bins
})



    











In [18]:



    


FIG_DATA_ACTIVITIES = "activities"

fig, ax = plt.subplots(figsize=figsize)
sns.barplot(x="activity", y="size", data=activities_df, ax=ax)
ax.set_xlabel("")
ax.set_ylabel("")
sns.despine()

plt.savefig(FIG_TEMPLATE.format(FIG_DATA_ACTIVITIES), bbox_inches="tight")
plt.savefig(FIG_HIGH_TEMPLATE.format(FIG_DATA_ACTIVITIES), bbox_inches="tight")



    


















    
























In [19]:



    


unique_subjects = features_df["subject"].unique()



    











In [20]:



    


activities_per_subject_df = pd.DataFrame(columns=["Subject", "Activity", "Size"])

for subject_id in unique_subjects:
    subject = str(subject_id)[2:4]
    subject_activities = labels_df[labels_df["subject"] == subject_id]["activity"].values
    subject_activities_bin = np.bincount(np.squeeze(subject_activities))
    
    for activity_id, activity in enumerate(ACTIVITIES):
        num_activity_samples = subject_activities_bin[activity_id]
        activities_per_subject_df.loc[len(activities_per_subject_df)] = [subject, activity, num_activity_samples]



    











In [21]:



    


FIG_DATA_ACTIVITIES_SUBJECT = "activities_per_subject"

g = sns.FacetGrid(data=activities_per_subject_df, col="Subject", col_wrap=4)
g = (g.map(sns.barplot, "Activity", "Size", palette=sns.color_palette(), ci=None)
     .set_xlabels("")
     .set_ylabels("")
     .set_xticklabels(rotation=90))
g.despine()
plt.yticks(np.arange(0, 2500, 500))

plt.savefig(FIG_TEMPLATE.format(FIG_DATA_ACTIVITIES_SUBJECT), bbox_inches="tight")
plt.savefig(FIG_HIGH_TEMPLATE.format(FIG_DATA_ACTIVITIES_SUBJECT), bbox_inches="tight")

UTILITY





In [22]:



    


def plot_evaluation(title, evaluation_results, best_iteration, best_score):
    # Create evaluation dataframe
    iterations = list(range(len(evaluation_results["eval"]["merror"]))) * 2
    num_iterations = len(iterations)
    errors = evaluation_results["eval"]["merror"] + evaluation_results["train"]["merror"]
    types = ["Validation"] * int(num_iterations/2) + ["Train"] * int(num_iterations/2)
    units = [0] * num_iterations
    data = {
        "Number of boosting rounds": iterations,
        "Mean multiclass classification error rate": errors,
        "Type": types,
        "unit": units
    }
    evaluation_dataframe = pd.DataFrame(data)
    
    # Plot
    fig, ax = plt.subplots(figsize=figsize)
    plt.plot([best_iteration], [best_score], 'o', markersize=20.0, markerfacecolor="None", markeredgecolor="green", markeredgewidth=3)

    sns.tsplot(data=evaluation_dataframe, color=sns.color_palette("Set1"), ax=ax,
               time="Number of boosting rounds", value="Mean multiclass classification error rate", 
               condition="Type", unit="unit")
    sns.despine()
    ax.set_title(title)
    ax.legend().title = ""
    
    # Anonotations
    ax.annotate("Best iteration (early-stopping): {}".format(best_iteration), xy=(0.05, 0.95), xycoords="axes fraction",
               fontsize=14)
    ax.annotate("Best score (early-stopping): {}".format(best_score), xy=(0.05, 0.90), xycoords="axes fraction",
               fontsize=14)

    return fig, ax



    











In [12]:



    


def plot_confusion_matrix(title, confusion_matrix):
    fig, ax = plt.subplots(figsize=figsize)
    sns.heatmap(data=confusion_matrix, annot=True, fmt=".2f", linewidths=1, square=True,
                vmin=0, vmax=100, ax=ax, xticklabels=ACTIVITIES, yticklabels=ACTIVITIES)
    plt.yticks(rotation=0)
    sns.despine()
#     ax.set_title(title)
    return fig, ax

LOAD RESULTS





In [24]:



    


with open("results/s1.pickle", "rb") as f:
    s1_results = pickle.load(f)

s1_eval_results = s1_results["eval_results"]
s1_eval_best_iteration = s1_results["eval_earlystoppping_best_iteration"]
s1_eval_best_score = s1_results["eval_earlystoppping_best_score"]
s1_accuracy = s1_results["final_accuracy"]
s1_confusion_matrix = s1_results["final_confusion_matrix"]



    











In [25]:



    


with open("results/s2.pickle", "rb") as f:
    s2_results = pickle.load(f)

s2_eval_results = s2_results["eval_results"]
s2_eval_best_iteration = s2_results["eval_earlystoppping_best_iteration"]
s2_eval_best_score = s2_results["eval_earlystoppping_best_score"]
s2_accuracy = s2_results["final_accuracy"]
s2_confusion_matrix = s2_results["final_confusion_matrix"]



    











In [2]:



    


with open("results/cs.pickle", "rb") as f:
    cs_results = pickle.load(f)

cs_training_time = cs_results["training_time"]
cs_testing_time = cs_results["testing_time"]
cs_eval_results = cs_results["eval_results"]
cs_eval_best_iteration = cs_results["eval_earlystoppping_best_iteration"]
cs_eval_best_score = cs_results["eval_earlystoppping_best_score"]
cs_accuracy = cs_results["final_accuracy"]
cs_training_accuracy = cs_results["training_accuracy"]
cs_confusion_matrix = cs_results["final_confusion_matrix"]
cs_classifier = cs_results["classifier"]
cs_confusion_matrix_subjects = cs_results["subject_confusion_matrix"]



    











In [27]:



    


with open("results/cs_rf.pickle", "rb") as f:
    cs_rf_results = pickle.load(f)

cs_rf_training_time = cs_rf_results["training_time"]
cs_rf_testing_time = cs_rf_results["testing_time"]
cs_rf_training_accuracy = cs_rf_results["training_accuracy"]
cs_rf_accuracy = cs_rf_results["final_accuracy"]
cs_rf_confusion_matrix = cs_rf_results["final_confusion_matrix"]



    











In [28]:



    


with open("results/cs_noinfrared.pickle", "rb") as f:
    noinfrared_results = pickle.load(f)

noinfrared_confusion_matrix = noinfrared_results["final_confusion_matrix"]



    











In [107]:



    


with open("results/cs_combinations.pickle", "rb") as f:
    combinations_results = pickle.load(f)

combinations_confusion_matrix = combinations_results["confusion_matrix"]

PLOT EVALUATION





In [81]:



    


fig, ax = plot_evaluation("Sample Test 1", s1_eval_results, s1_eval_best_iteration, s1_eval_best_score)
FIG_S1_EVAL_ITERATIONS = "s1_iterations"
fig.savefig(FIG_TEMPLATE.format(FIG_S1_EVAL_ITERATIONS), bbox_inches="tight")
fig.savefig(FIG_HIGH_TEMPLATE.format(FIG_S1_EVAL_ITERATIONS), bbox_inches="tight")



    


















    
























In [82]:



    


fig, ax = plot_evaluation("Sample Test 2", s2_eval_results, s2_eval_best_iteration, s2_eval_best_score)
FIG_S2_EVAL_ITERATIONS = "s2_evaluation_iterations"
fig.savefig(FIG_TEMPLATE.format(FIG_S2_EVAL_ITERATIONS), bbox_inches="tight")
fig.savefig(FIG_HIGH_TEMPLATE.format(FIG_S2_EVAL_ITERATIONS), bbox_inches="tight")



    


















    
























In [83]:



    


fig, ax = plot_evaluation("Cross-Subject Test", cs_eval_results, cs_eval_best_iteration, cs_eval_best_score)
FIG_CS_EVAL_ITERATIONS = "cs_evaluation_iterations"
fig.savefig(FIG_TEMPLATE.format(FIG_CS_EVAL_ITERATIONS), bbox_inches="tight")
fig.savefig(FIG_HIGH_TEMPLATE.format(FIG_CS_EVAL_ITERATIONS), bbox_inches="tight")

PLOT CONFUSION MATRIX





In [84]:



    


fig, ax = plot_confusion_matrix("Samples Test 1", s1_confusion_matrix)
FIG_S1_CONFUSION_MATRIX = "s1_confusion_matrix"
fig.savefig(FIG_TEMPLATE.format(FIG_S1_CONFUSION_MATRIX), bbox_inches="tight")
fig.savefig(FIG_HIGH_TEMPLATE.format(FIG_S1_CONFUSION_MATRIX), bbox_inches="tight")



    


















    
























In [85]:



    


fig, ax = plot_confusion_matrix("Samples Test 2", s2_confusion_matrix)
FIG_S2_CONFUSION_MATRIX = "s2_confusion_matrix"
fig.savefig(FIG_TEMPLATE.format(FIG_S2_CONFUSION_MATRIX), bbox_inches="tight")
fig.savefig(FIG_HIGH_TEMPLATE.format(FIG_S2_CONFUSION_MATRIX), bbox_inches="tight")



    


















    
























In [86]:



    


fig, ax = plot_confusion_matrix("Cross-Subject Test", cs_confusion_matrix)
FIG_CS_CONFUSION_MATRIX = "cs_confusion_matrix"
fig.savefig(FIG_TEMPLATE.format(FIG_CS_CONFUSION_MATRIX), bbox_inches="tight")
fig.savefig(FIG_HIGH_TEMPLATE.format(FIG_CS_CONFUSION_MATRIX), bbox_inches="tight")



    


















    
























In [ ]:



    


# paper figs



    











In [5]:



    


ACTIVITIES = ["Standing", "Sitting", "Pointing", "Phone", "Tablet", "Paper"]



    











In [6]:



    


sns.set()
sns.set(context="notebook", style="ticks", font_scale=1.8)
figsize=(10, 7.5)



    











In [37]:



    


init_cs = pd.read_csv("Q:\\p_eaglesense\\CHI\\figs\\Figure 7 left.csv", header=None)



    











In [38]:



    


init_cs



    


















    
Out[38]:










  
    

      
      0
      1
      2
      3
      4
      5
    

  
  
    

      0
      96.56
      0.13
      0.00
      1.17
      1.93
      0.21
    

    

      1
      0.54
      98.24
      0.60
      0.02
      0.56
      0.01
    

    

      2
      0.34
      6.42
      92.99
      0.02
      0.15
      0.08
    

    

      3
      32.97
      0.00
      0.00
      63.82
      2.83
      0.39
    

    

      4
      1.20
      0.07
      0.00
      3.76
      93.18
      1.80
    

    

      5
      6.89
      0.40
      0.01
      0.72
      0.24
      91.73
    

  




















In [84]:



    


fig, ax = plt.subplots(figsize=figsize)
sns.heatmap(data=init_cs, annot=True, fmt=".2f", linewidths=1, square=True,
            vmin=0, vmax=100, ax=ax, xticklabels=ACTIVITIES, yticklabels=ACTIVITIES)
plt.yticks(rotation=0)
sns.despine()
fig.savefig("Q:\\p_eaglesense\\CHI\\figs\\Figure 7 left-600dpi.png", bbox_inches="tight", dpi=600)



    


















    
























In [86]:



    


complete_cs = pd.read_csv("Q:\\p_eaglesense\\CHI\\figs\\Figure 7 right.csv", header=None)



    











In [87]:



    


complete_cs



    


















    
Out[87]:










  
    

      
      0
      1
      2
      3
      4
      5
    

  
  
    

      0
      96.70
      0.14
      0.02
      1.24
      1.21
      0.68
    

    

      1
      0.52
      96.33
      2.76
      0.02
      0.27
      0.10
    

    

      2
      2.89
      4.19
      92.61
      0.03
      0.19
      0.10
    

    

      3
      22.46
      0.02
      0.01
      71.59
      2.67
      3.24
    

    

      4
      5.54
      0.10
      0.01
      9.48
      83.30
      1.56
    

    

      5
      3.91
      0.40
      0.09
      2.21
      0.57
      92.81
    

  




















In [88]:



    


fig, ax = plt.subplots(figsize=figsize)
sns.heatmap(data=complete_cs, annot=True, fmt=".2f", linewidths=1, square=True,
            vmin=0, vmax=100, ax=ax, xticklabels=ACTIVITIES, yticklabels=ACTIVITIES)
plt.yticks(rotation=0)
sns.despine()
fig.savefig("Q:\\p_eaglesense\\CHI\\figs\\Figure 7 right-600dpi.png", bbox_inches="tight", dpi=600)



    


















    
























In [63]:



    


infrared_compared = pd.read_csv("Q:\\p_eaglesense\\CHI\\figs\\Figure 8.csv", header=None)



    











In [64]:



    


infrared_compared



    


















    
Out[64]:










  
    

      
      0
      1
    

  
  
    

      0
      96.56
      98.01
    

    

      1
      98.24
      98.47
    

    

      2
      92.99
      93.18
    

    

      3
      63.84
      53.55
    

    

      4
      93.18
      60.06
    

    

      5
      91.73
      86.53
    

  




















In [65]:



    


infrared_compared_df = pd.DataFrame(columns=["Activity", "Accuracy", "Condition"])
for idx, activity in enumerate(ACTIVITIES):
    infrared_accuracy = infrared_compared.loc[idx, 0]
    noinfrared_accuracy = infrared_compared.loc[idx, 1]
    infrared_compared_df.loc[len(infrared_compared_df)] = [
        activity, infrared_accuracy, "All"
    ]
    infrared_compared_df.loc[len(infrared_compared_df)] = [
        activity, noinfrared_accuracy, "No Infrareds"
    ]



    











In [66]:



    


color = sns.color_palette()



    











In [89]:



    


fig, ax = plt.subplots(figsize=figsize)
sns.barplot(data=infrared_compared_df, ax=ax,
            x="Activity", y="Accuracy", hue="Condition", palette=[color[0], color[2]])
sns.despine()
ax.set_xlabel("")
ax.set_ylabel("")
ax.legend(loc="upper center", bbox_to_anchor=(0.5, -0.08), frameon=True, framealpha=1, fancybox=True, shadow=True, ncol=2)
for p in ax.patches:
    height = p.get_height()
    ax.text(p.get_x()+0.06, height+2, "%1.0f"%(height))
fig.savefig("Q:\\p_eaglesense\\CHI\\figs\\Figure 8-600dpi.png", bbox_inches="tight", dpi=600)



    


















    
























In [13]:



    


np.set_printoptions(formatter={'float': lambda x: "{:.2f}".format(x)})



    











In [18]:



    


cs_confusion_matrix_subjects



    


















    
Out[18]:








[(2001.0, 0.9980318482733942, array([[100.00, 0.00, 0.00, 0.00, 0.00, 0.00],
         [0.00, 99.93, 0.07, 0.00, 0.00, 0.00],
         [0.00, 1.06, 98.94, 0.00, 0.00, 0.00],
         [0.49, 0.00, 0.00, 99.51, 0.00, 0.00],
         [0.00, 0.00, 0.00, 0.00, 100.00, 0.00],
         [0.00, 0.00, 0.00, 0.00, 0.00, 100.00]])),
 (2002.0, 0.90640038498556308, array([[96.19, 0.61, 0.00, 0.76, 1.98, 0.46],
         [0.71, 99.22, 0.06, 0.00, 0.00, 0.00],
         [2.84, 10.28, 84.75, 0.00, 1.77, 0.35],
         [11.37, 0.00, 0.00, 58.60, 25.95, 4.08],
         [1.41, 0.12, 0.00, 0.00, 84.29, 14.19],
         [4.78, 1.66, 0.00, 0.00, 0.42, 93.14]])),
 (2003.0, 0.99919175591028486, array([[99.87, 0.00, 0.00, 0.00, 0.13, 0.00],
         [0.00, 100.00, 0.00, 0.00, 0.00, 0.00],
         [0.00, 0.45, 99.55, 0.00, 0.00, 0.00],
         [0.00, 0.00, 0.00, 100.00, 0.00, 0.00],
         [0.13, 0.00, 0.00, 0.00, 99.74, 0.13],
         [0.00, 0.00, 0.00, 0.00, 0.00, 100.00]])),
 (2004.0, 0.97154072620215903, array([[92.64, 0.12, 0.00, 4.83, 2.42, 0.00],
         [0.08, 97.86, 0.08, 0.00, 1.98, 0.00],
         [0.00, 3.57, 96.26, 0.00, 0.17, 0.00],
         [0.13, 0.00, 0.00, 99.22, 0.65, 0.00],
         [1.15, 0.00, 0.00, 0.35, 98.50, 0.00],
         [1.76, 0.00, 0.00, 0.00, 0.00, 98.24]])),
 (2005.0, 0.99964074007544457, array([[99.90, 0.00, 0.00, 0.10, 0.00, 0.00],
         [0.00, 100.00, 0.00, 0.00, 0.00, 0.00],
         [0.00, 0.25, 99.75, 0.00, 0.00, 0.00],
         [0.00, 0.00, 0.00, 100.00, 0.00, 0.00],
         [0.00, 0.00, 0.00, 0.00, 100.00, 0.00],
         [0.00, 0.00, 0.00, 0.00, 0.00, 100.00]])),
 (2006.0, 0.89053553384978112, array([[98.50, 0.00, 0.00, 0.12, 1.12, 0.25],
         [1.61, 96.46, 0.48, 0.00, 1.45, 0.00],
         [0.00, 1.77, 98.15, 0.00, 0.08, 0.00],
         [38.74, 0.00, 0.00, 59.85, 1.42, 0.00],
         [2.85, 0.20, 0.00, 0.49, 96.17, 0.29],
         [20.40, 0.00, 0.00, 4.55, 0.70, 74.36]])),
 (2007.0, 0.99934454883111212, array([[100.00, 0.00, 0.00, 0.00, 0.00, 0.00],
         [0.00, 99.93, 0.07, 0.00, 0.00, 0.00],
         [0.00, 0.46, 99.54, 0.00, 0.00, 0.00],
         [0.00, 0.00, 0.00, 100.00, 0.00, 0.00],
         [0.00, 0.00, 0.00, 0.00, 100.00, 0.00],
         [0.00, 0.00, 0.00, 0.00, 0.00, 100.00]])),
 (2008.0, 0.96580003489792354, array([[96.96, 0.09, 0.00, 1.84, 0.65, 0.46],
         [0.31, 97.77, 1.77, 0.15, 0.00, 0.00],
         [0.52, 0.26, 98.69, 0.13, 0.00, 0.39],
         [5.73, 0.00, 0.00, 92.83, 1.08, 0.36],
         [1.22, 0.00, 0.00, 2.24, 96.03, 0.51],
         [3.28, 0.00, 0.00, 0.00, 0.00, 96.72]])),
 (2009.0, 0.99894358757658985, array([[99.86, 0.00, 0.00, 0.14, 0.00, 0.00],
         [0.09, 99.91, 0.00, 0.00, 0.00, 0.00],
         [0.00, 0.11, 99.89, 0.00, 0.00, 0.00],
         [0.00, 0.00, 0.00, 100.00, 0.00, 0.00],
         [0.15, 0.00, 0.00, 0.15, 99.70, 0.00],
         [0.00, 0.00, 0.00, 0.00, 0.00, 100.00]])),
 (2010.0, 0.8858234691024327, array([[98.56, 0.00, 0.00, 0.63, 0.72, 0.09],
         [0.58, 99.02, 0.31, 0.00, 0.09, 0.00],
         [0.20, 14.92, 84.88, 0.00, 0.00, 0.00],
         [53.25, 0.00, 0.00, 43.92, 2.77, 0.06],
         [0.94, 0.00, 0.00, 0.04, 99.02, 0.00],
         [5.73, 0.52, 0.06, 0.46, 0.35, 92.88]])),
 (2011.0, 0.99926547743966421, array([[100.00, 0.00, 0.00, 0.00, 0.00, 0.00],
         [0.00, 99.96, 0.00, 0.00, 0.04, 0.00],
         [0.00, 0.49, 99.51, 0.00, 0.00, 0.00],
         [0.00, 0.00, 0.00, 100.00, 0.00, 0.00],
         [0.00, 0.00, 0.00, 0.00, 100.00, 0.00],
         [0.07, 0.00, 0.00, 0.00, 0.07, 99.87]])),
 (2012.0, 0.86836049856184083, array([[94.75, 0.20, 0.00, 0.13, 4.72, 0.20],
         [0.28, 98.25, 0.97, 0.00, 0.46, 0.05],
         [0.22, 11.38, 88.40, 0.00, 0.00, 0.00],
         [44.32, 0.00, 0.00, 54.94, 0.41, 0.34],
         [0.34, 0.14, 0.00, 17.31, 81.80, 0.41],
         [6.05, 0.46, 0.00, 0.12, 0.12, 93.24]]))]

















In [17]:



    


for subject_id, _, subject_confusion_matrix in cs_confusion_matrix_subjects:
    subject = str(subject_id)[2:4]
    print("Subject", subject)
    print(subject_confusion_matrix)



    


















    






Subject 01
[[100.00 0.00 0.00 0.00 0.00 0.00]
 [0.00 99.93 0.07 0.00 0.00 0.00]
 [0.00 1.06 98.94 0.00 0.00 0.00]
 [0.49 0.00 0.00 99.51 0.00 0.00]
 [0.00 0.00 0.00 0.00 100.00 0.00]
 [0.00 0.00 0.00 0.00 0.00 100.00]]
Subject 02
[[96.19 0.61 0.00 0.76 1.98 0.46]
 [0.71 99.22 0.06 0.00 0.00 0.00]
 [2.84 10.28 84.75 0.00 1.77 0.35]
 [11.37 0.00 0.00 58.60 25.95 4.08]
 [1.41 0.12 0.00 0.00 84.29 14.19]
 [4.78 1.66 0.00 0.00 0.42 93.14]]
Subject 03
[[99.87 0.00 0.00 0.00 0.13 0.00]
 [0.00 100.00 0.00 0.00 0.00 0.00]
 [0.00 0.45 99.55 0.00 0.00 0.00]
 [0.00 0.00 0.00 100.00 0.00 0.00]
 [0.13 0.00 0.00 0.00 99.74 0.13]
 [0.00 0.00 0.00 0.00 0.00 100.00]]
Subject 04
[[92.64 0.12 0.00 4.83 2.42 0.00]
 [0.08 97.86 0.08 0.00 1.98 0.00]
 [0.00 3.57 96.26 0.00 0.17 0.00]
 [0.13 0.00 0.00 99.22 0.65 0.00]
 [1.15 0.00 0.00 0.35 98.50 0.00]
 [1.76 0.00 0.00 0.00 0.00 98.24]]
Subject 05
[[99.90 0.00 0.00 0.10 0.00 0.00]
 [0.00 100.00 0.00 0.00 0.00 0.00]
 [0.00 0.25 99.75 0.00 0.00 0.00]
 [0.00 0.00 0.00 100.00 0.00 0.00]
 [0.00 0.00 0.00 0.00 100.00 0.00]
 [0.00 0.00 0.00 0.00 0.00 100.00]]
Subject 06
[[98.50 0.00 0.00 0.12 1.12 0.25]
 [1.61 96.46 0.48 0.00 1.45 0.00]
 [0.00 1.77 98.15 0.00 0.08 0.00]
 [38.74 0.00 0.00 59.85 1.42 0.00]
 [2.85 0.20 0.00 0.49 96.17 0.29]
 [20.40 0.00 0.00 4.55 0.70 74.36]]
Subject 07
[[100.00 0.00 0.00 0.00 0.00 0.00]
 [0.00 99.93 0.07 0.00 0.00 0.00]
 [0.00 0.46 99.54 0.00 0.00 0.00]
 [0.00 0.00 0.00 100.00 0.00 0.00]
 [0.00 0.00 0.00 0.00 100.00 0.00]
 [0.00 0.00 0.00 0.00 0.00 100.00]]
Subject 08
[[96.96 0.09 0.00 1.84 0.65 0.46]
 [0.31 97.77 1.77 0.15 0.00 0.00]
 [0.52 0.26 98.69 0.13 0.00 0.39]
 [5.73 0.00 0.00 92.83 1.08 0.36]
 [1.22 0.00 0.00 2.24 96.03 0.51]
 [3.28 0.00 0.00 0.00 0.00 96.72]]
Subject 09
[[99.86 0.00 0.00 0.14 0.00 0.00]
 [0.09 99.91 0.00 0.00 0.00 0.00]
 [0.00 0.11 99.89 0.00 0.00 0.00]
 [0.00 0.00 0.00 100.00 0.00 0.00]
 [0.15 0.00 0.00 0.15 99.70 0.00]
 [0.00 0.00 0.00 0.00 0.00 100.00]]
Subject 10
[[98.56 0.00 0.00 0.63 0.72 0.09]
 [0.58 99.02 0.31 0.00 0.09 0.00]
 [0.20 14.92 84.88 0.00 0.00 0.00]
 [53.25 0.00 0.00 43.92 2.77 0.06]
 [0.94 0.00 0.00 0.04 99.02 0.00]
 [5.73 0.52 0.06 0.46 0.35 92.88]]
Subject 11
[[100.00 0.00 0.00 0.00 0.00 0.00]
 [0.00 99.96 0.00 0.00 0.04 0.00]
 [0.00 0.49 99.51 0.00 0.00 0.00]
 [0.00 0.00 0.00 100.00 0.00 0.00]
 [0.00 0.00 0.00 0.00 100.00 0.00]
 [0.07 0.00 0.00 0.00 0.07 99.87]]
Subject 12
[[94.75 0.20 0.00 0.13 4.72 0.20]
 [0.28 98.25 0.97 0.00 0.46 0.05]
 [0.22 11.38 88.40 0.00 0.00 0.00]
 [44.32 0.00 0.00 54.94 0.41 0.34]
 [0.34 0.14 0.00 17.31 81.80 0.41]
 [6.05 0.46 0.00 0.12 0.12 93.24]]

















In [14]:



    


even_subject_cs_df = pd.DataFrame(columns=["Subject", "Activity", "Accuracy"])
for subject_id, _, subject_confusion_matrix in cs_confusion_matrix_subjects:
    subject = str(subject_id)[2:4]
    for activity_id, activity in enumerate(ACTIVITIES):
        activity_accuracy = subject_confusion_matrix[activity_id, activity_id]
        if subject_id % 2 == 0:
            even_subject_cs_df.loc[len(cs_cofnusion_matrix_subject_df)] = [subject, activity, activity_accuracy]



    











In [58]:



    


g = sns.FacetGrid(data=cs_cofnusion_matrix_subject_df, col="Subject", col_wrap=3)
g = (g.map(sns.barplot, "Activity", "Accuracy", palette=sns.color_palette(), ci=None)
     .set_xlabels("")
     .set_ylabels("")
     .set_xticklabels(rotation=90))
g.despine()
plt.yticks(np.arange(0, 101, 20))
g.fig.savefig("Q:\\p_eaglesense\\CHI\\figs\\even-subject-cs-1000dpi.png", bbox_inches="tight", dpi=1000)



    


















    
























In [46]:



    


all_subject_cs_df = pd.DataFrame(columns=["Subject", "Activity", "Accuracy"])
for subject_id, _, subject_confusion_matrix in cs_confusion_matrix_subjects:
    subject = str(subject_id)[2:4]
    for activity_id, activity in enumerate(ACTIVITIES):
        activity_accuracy = subject_confusion_matrix[activity_id, activity_id]
        all_subject_cs_df.loc[len(all_subject_cs_df)] = [subject, activity, activity_accuracy]



    











In [56]:



    


g = sns.FacetGrid(data=all_subject_cs_df, col="Subject", col_wrap=4)
g = (g.map(sns.barplot, "Activity", "Accuracy", palette=sns.color_palette(), ci=None)
     .set_xlabels("")
     .set_ylabels("")
     .set_xticklabels(rotation=90))
g.despine()
plt.yticks(np.arange(0, 101, 20))
g.fig.savefig("Q:\\p_eaglesense\\CHI\\figs\\all-subject-cs-1000dpi.png", bbox_inches="tight", dpi=1000)



    


















    
























In [ ]:

CROSS SUBJECT TEST





In [87]:



    


feature_importance = cs_classifier.get_fscore()



    











In [88]:



    


# Create feature importance dataframe
features_importance_formatted = dict()
for feature_idx, feature_name in enumerate(features_vector.columns):
    old_key = "f{}".format(feature_idx)
    if old_key not in feature_importance:
        continue
    new_key = feature_name
    features_importance_formatted[new_key] = feature_importance[old_key]
features_importance_formatted = sorted(features_importance_formatted.items(), key=operator.itemgetter(1))
features_importance_top10 = features_importance_formatted[-10:]
features_importance_df = pd.DataFrame(features_importance_top10, columns=["feature", "gain"])
features_importance_df["gain"] = features_importance_df["gain"] / features_importance_df["gain"].sum()

# Plot
fig, ax = plt.subplots(figsize=figsize)
sns.barplot(x="gain", y="feature", data=features_importance_df, label="Total", color="#3498db", ax=ax)
sns.despine()
ax.set_xlabel("Gain")
ax.set_ylabel("Feature")
ax.set_title("Cross-Subject Test")

FIG_CS_FEATURE_IMPORTANCE = "cs_feature_importance"
fig.savefig(FIG_TEMPLATE.format(FIG_CS_FEATURE_IMPORTANCE), bbox_inches="tight")
fig.savefig(FIG_HIGH_TEMPLATE.format(FIG_CS_FEATURE_IMPORTANCE), bbox_inches="tight")



    


















    
























In [89]:



    


np.set_printoptions(formatter={'float': lambda x: "{:.2f}".format(x)})



    











In [124]:



    


cs_cofnusion_matrix_subject_df = pd.DataFrame(columns=["Subject", "Activity", "Accuracy"])
for subject_id, _, subject_confusion_matrix in cs_confusion_matrix_subjects:
    subject = str(subject_id)[2:4]
    for activity_id, activity in enumerate(ACTIVITIES):
        activity_accuracy = subject_confusion_matrix[activity_id, activity_id]
        if subject_id % 2 == 0:
            cs_cofnusion_matrix_subject_df.loc[len(cs_cofnusion_matrix_subject_df)] = [subject, activity, activity_accuracy]
        if activity == "Phone":
            print("Subject", subject)
            print(subject_confusion_matrix[activity_id,:])



    


















    






Subject 01
[0.49 0.00 0.00 99.51 0.00 0.00]
Subject 02
[11.37 0.00 0.00 58.60 25.95 4.08]
Subject 03
[0.00 0.00 0.00 100.00 0.00 0.00]
Subject 04
[0.13 0.00 0.00 99.22 0.65 0.00]
Subject 05
[0.00 0.00 0.00 100.00 0.00 0.00]
Subject 06
[38.74 0.00 0.00 59.85 1.42 0.00]
Subject 07
[0.00 0.00 0.00 100.00 0.00 0.00]
Subject 08
[5.73 0.00 0.00 92.83 1.08 0.36]
Subject 09
[0.00 0.00 0.00 100.00 0.00 0.00]
Subject 10
[53.25 0.00 0.00 43.92 2.77 0.06]
Subject 11
[0.00 0.00 0.00 100.00 0.00 0.00]
Subject 12
[44.32 0.00 0.00 54.94 0.41 0.34]

















In [40]:



    


g = sns.FacetGrid(data=cs_cofnusion_matrix_subject_df, col="Subject", col_wrap=3)
g = (g.map(sns.barplot, "Activity", "Accuracy", palette=sns.color_palette(), ci=None)
     .set_xlabels("")
     .set_ylabels("")
     .set_xticklabels(rotation=90))
g.despine()
plt.yticks(np.arange(0, 101, 20))

FIG_CS_CONFUSION_MATRIX_SUBJECT = "cs_accuracy_subject"
g.fig.savefig(FIG_TEMPLATE.format(FIG_CS_CONFUSION_MATRIX_SUBJECT), bbox_inches="tight")
g.fig.savefig(FIG_HIGH_TEMPLATE.format(FIG_CS_CONFUSION_MATRIX_SUBJECT), bbox_inches="tight")



    


















    






---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
<ipython-input-40-141e9312c6f1> in <module>()
      8 
      9 FIG_CS_CONFUSION_MATRIX_SUBJECT = "cs_accuracy_subject"
---> 10 g.fig.savefig(FIG_TEMPLATE.format(FIG_CS_CONFUSION_MATRIX_SUBJECT), bbox_inches="tight")
     11 g.fig.savefig(FIG_HIGH_TEMPLATE.format(FIG_CS_CONFUSION_MATRIX_SUBJECT), bbox_inches="tight")

NameError: name 'FIG_TEMPLATE' is not defined









    
























In [125]:



    


g = sns.FacetGrid(data=cs_cofnusion_matrix_subject_df, col="Subject", col_wrap=3)
g = (g.map(sns.barplot, "Activity", "Accuracy", palette=sns.color_palette(), ci=None)
     .set_xlabels("")
     .set_ylabels("")
     .set_xticklabels(rotation=90))
g.despine()
plt.yticks(np.arange(0, 101, 20))

FIG_CS_CONFUSION_MATRIX_SUBJECT = "cs_accuracy_subject"
g.fig.savefig(FIG_TEMPLATE.format(FIG_CS_CONFUSION_MATRIX_SUBJECT), bbox_inches="tight")
g.fig.savefig(FIG_HIGH_TEMPLATE.format(FIG_CS_CONFUSION_MATRIX_SUBJECT), bbox_inches="tight")

Random Forest





In [92]:



    


rf_comparison_data = [
    {
        "Test": "Train",
        "Classifier": "Random Forest",
        "Time": cs_rf_training_time,
        "Accuracy": cs_rf_training_accuracy
    },
    {
        "Test": "Test",
        "Classifier": "Random Forest",
        "Time": cs_rf_testing_time,
        "Accuracy": cs_rf_accuracy
    },
    {
        "Test": "Train",
        "Classifier": "XGBoost",
        "Time": cs_training_time,
        "Accuracy": cs_training_accuracy
    },
    {
        "Test": "Test",
        "Classifier": "XGBoost",
        "Time": cs_testing_time,
        "Accuracy": cs_accuracy
    },
]

rf_comparison_df = pd.DataFrame(rf_comparison_data, columns=["Test", "Classifier", "Time", "Accuracy"])
rf_comparison_df.head()



    


















    
Out[92]:










  
    

      
      Test
      Classifier
      Time
      Accuracy
    

  
  
    

      0
      Train
      Random Forest
      2.524389
      1.000000
    

    

      1
      Test
      Random Forest
      0.270627
      0.844388
    

    

      2
      Train
      XGBoost
      9.322780
      1.000000
    

    

      3
      Test
      XGBoost
      0.231154
      0.905464
    

  




















In [93]:



    


fig, ax = plt.subplots(figsize=figsize)
sns.barplot(data=rf_comparison_df, x="Test", y="Time", hue="Classifier", ax=ax)
ax.set_xlabel("")
ax.set_ylabel("Time")
sns.despine()

FIG_CS_RANDOMFOREST_TIME = "cs_randomforest_time"
fig.savefig(FIG_TEMPLATE.format(FIG_CS_RANDOMFOREST_TIME), bbox_inches="tight")
fig.savefig(FIG_HIGH_TEMPLATE.format(FIG_CS_RANDOMFOREST_TIME), bbox_inches="tight")



    


















    
























In [94]:



    


fig, ax = plt.subplots(figsize=figsize)
sns.barplot(data=rf_comparison_df, x="Test", y="Accuracy", hue="Classifier", ax=ax)
ax.set_xlabel("")
ax.set_ylabel("Accuracy")
sns.despine()

FIG_CS_RANDOMFOREST_ACCURACY = "cs_randomforest_accuracy"
fig.savefig(FIG_TEMPLATE.format(FIG_CS_RANDOMFOREST_ACCURACY), bbox_inches="tight")
fig.savefig(FIG_HIGH_TEMPLATE.format(FIG_CS_RANDOMFOREST_ACCURACY), bbox_inches="tight")

No infrared





In [95]:



    


fig, ax = plot_confusion_matrix("Cross-Subject Test (No infrareds)", noinfrared_confusion_matrix)
FIG_CS_CONFUSION_MATRIX_NONINFRARED = "cs_noinfrared_confusion_matrix"
fig.savefig(FIG_TEMPLATE.format(FIG_CS_CONFUSION_MATRIX_NONINFRARED), bbox_inches="tight")
fig.savefig(FIG_HIGH_TEMPLATE.format(FIG_CS_CONFUSION_MATRIX_NONINFRARED), bbox_inches="tight")



    


















    
























In [34]:



    


crosssubject_infrared_comparisons_df = pd.DataFrame(columns=["Activity", "Accuracy", "Condition"])

for activity_id, activity in enumerate(ACTIVITIES):
    infrared_accuracy = cs_confusion_matrix[activity_id, activity_id]
    noinfrared_accuracy = noinfrared_confusion_matrix[activity_id, activity_id]
    
    crosssubject_infrared_comparisons_df.loc[len(crosssubject_infrared_comparisons_df)] = [
        activity, infrared_accuracy, "All"
    ]
    crosssubject_infrared_comparisons_df.loc[len(crosssubject_infrared_comparisons_df)] = [
        activity, noinfrared_accuracy, "No infrareds"
    ]
    
    print(activity, infrared_accuracy - noinfrared_accuracy)



    


















    






Standing -1.45393541172
Sitting -0.235607457488
Pointing -0.188956539996
Phone 10.2659664577
Tablet 33.1205107741
Paper 5.2005730659

















In [97]:



    


sns.palplot(sns.color_palette("coolwarm", n_colors=7))



    


















    
























In [98]:



    


sns.palplot(sns.color_palette("hls", 8))



    


















    
























In [99]:



    


sns.palplot(sns.color_palette("Set2", 10))



    


















    
























In [100]:



    


flatui = ["#9b59b6", "#3498db", "#95a5a6", "#e74c3c", "#34495e", "#2ecc71"]



    











In [101]:



    


sns.palplot(sns.color_palette(flatui))



    


















    
























In [31]:



    


sns.palplot(sns.color_palette())



    


















    
























In [32]:



    


default_colors = sns.color_palette()



    











In [36]:



    


fig, ax = plt.subplots(figsize=figsize)
sns.barplot(data=crosssubject_infrared_comparisons_df, ax=ax,
            x="Activity", y="Accuracy", hue="Condition", palette=[default_colors[0], default_colors[2]])
sns.despine()
ax.set_xlabel("")
ax.set_ylabel("")
ax.legend(loc='upper center', bbox_to_anchor=(0.5, -0.05), frameon=True, framealpha=1, fancybox=True, shadow=True, ncol=2)
# ax.set_title("Cross-Subject Test")

for p in ax.patches:
    height = p.get_height()
    ax.text(p.get_x(), height+ 3, '%1.2f'%(height/100))

    
FIG_CS_INFRARED_COMPARISON = "cs_infrared_comparison"
fig.savefig(FIG_TEMPLATE.format(FIG_CS_INFRARED_COMPARISON), bbox_inches="tight")
fig.savefig(FIG_HIGH_TEMPLATE.format(FIG_CS_INFRARED_COMPARISON), bbox_inches="tight")



    


















    
























In [120]:



    


fig, ax = plot_confusion_matrix("Cross-Subject Test (All Combinations)", combinations_confusion_matrix)
FIG_CS_CONFUSION_MATRIX_ALL = "cs_all_confusion_matrix"
fig.savefig(FIG_TEMPLATE.format(FIG_CS_CONFUSION_MATRIX_ALL), bbox_inches="tight")
fig.savefig(FIG_HIGH_TEMPLATE.format(FIG_CS_CONFUSION_MATRIX_ALL), bbox_inches="tight")



    


















    
























In [53]:



    


fig, ax = plt.subplots(figsize=(100,100))
xgb.plot_tree(cs_classifier, num_trees=0, rankdir="UT", ax=ax)
FIG_CS_CLASSIFIER = "cs_classifier"
fig.savefig(FIG_TEMPLATE.format(FIG_CS_CLASSIFIER), bbox_inches="tight")
fig.savefig(FIG_HIGH_TEMPLATE.format(FIG_CS_CLASSIFIER), bbox_inches="tight")

Content source: cjw-charleswu/eaglesense

Similar notebooks:

notebook.community | gallery | about