def get_behavior_df(date, cow_id, model1, model2):
""" 個々の牛の1日分の行動分類を行う """
print(sys._getframe().f_code.co_name, "実行中")
print("行動分類を行います---：日付: ", date.strftime("%Y/%m/%d"), ", 個体番号: ", cow_id)
filename = "behavior_classification/training_data/features.csv"
existed = output_features.output_features(filename, date, cow_id)
if (existed): # GPSデータが存在していれば
### 各特徴から行動を分類する ###
df = pd.read_csv(filename, sep = ",", header = 0, usecols = [0,1,2,3,4,5,6,7,9], names=('Time', 'RCategory', 'WCategory', 'RTime', 'WTime', 'AccumulatedDis', 'Velocity', 'MVelocity', 'Distance')) # csv読み込み
labels = []
probs = []
x1, x2, x3, x4, x5, x6, x7, x8 = df['RCategory'].tolist(), df['WCategory'].tolist(), df['RTime'].tolist(), df['WTime'].tolist(), df['AccumulatedDis'].tolist(), df['Velocity'].tolist(), df['MVelocity'].tolist(), df['Distance'].tolist()
x = np.array((x1, x2, x3, x4, x5, x6, x7, x8)).T
result1, prob1 = model1.predict(x), model1.predict_proba(x) # 休息セグメントの結果と確率
result2, prob2 = model2.predict(x), model2.predict_proba(x) # 活動セグメントの結果と確率
for a, b, c, d in zip(result1, result2, prob1, prob2):
labels.append(a)
labels.append(b)
probs.append(np.insert(c, 2, 0.0))
probs.append(d)
# --- 復元 ---
t_list = loading.make_time_list(date)
zipped_t_list = regex.str_to_datetime(df['Time'].tolist())
times, labels = postprocessing.decompress(t_list, zipped_t_list, labels)
df = cut_used_time(times, labels, cow_id)
return df
else:
t_list = loading.make_time_list(date)
labels = [-1 for l in range(len(t_list))]
df = cut_used_time(t_list, labels, cow_id)
return df
), df['WTime'].tolist(), df['AccumulatedDis'].tolist(
), df['Velocity'].tolist(), df['MVelocity'].tolist(
), df['Distance'].tolist()
x = np.array((x1, x2, x3, x4, x5, x6, x7, x8)).T
result1 = model1.predict(x)
result2 = model2.predict(x)
prob1 = model1.predict_proba(x)
prob2 = model2.predict_proba(x)
print(result1)
print(result2)
for a, b, c, d in zip(result1, result2, prob1, prob2):
if (c.max() >= 0.7 and a == 1):
labels.append(a)
else:
labels.append(0)
# --- 復元 ---
zipped_t_list = regex.str_to_datetime(df['Time'].tolist())
new_t_list, labels = postprocessing.decompress(
t_list, zipped_t_list, labels)
sum_rest = 0
for i in labels:
if (i == 0):
sum_rest += 5
sum_rest_list.append(sum_rest)
else:
sum_rest_list.append(0)
pd.DataFrame(date_list).to_csv("date.csv")
pd.DataFrame(sum_rest_list).to_csv("testa.csv")