def plotDay(config, timestamps, realY, predictY):
assert len(realY) == len(predictY)
realMeans, realSd = getMeanSdDay(config, realY)
predictedMeans, predictedSd = getMeanSdDay(config, predictY)
x1 = list(range(96))
plt.plot(x1, realMeans, label="actual", color="green")
plt.fill_between(
x1, realMeans - realSd * 0.5, realMeans + realSd * 0.5, color="green", alpha=0.5
)
plt.plot(x1, predictedMeans, label="predict", color="orange")
plt.fill_between(
x1,
predictedMeans - predictedSd * 0.5,
predictedMeans + predictedSd * 0.5,
color="orange",
alpha=0.5,
)
time, tick = makeTick(timestamps[:96], "%H:%M")
plt.xticks(tick, time, rotation=20)
plt.xlabel("Time of Day")
plt.ylabel("Average Power consumption (kW)")
plt.legend()
plt.tight_layout()
plt.show()
def plotEcart(
train_y,
train_predict_y,
val_y,
val_predict_y,
test_y,
test_predict_y,
timestamps,
config,
):
time, tick = makeTick(timestamps)
y1, y1b = [], []
for i in range(
min((len(train_predict_y) // config.OUTPUT_SIZE) - 1, config.NB_PLOT)
):
y1.extend(train_y[i * config.OUTPUT_SIZE])
y1b.extend(train_predict_y[i * config.OUTPUT_SIZE])
y2, y2b = [], []
for i in range(min((len(val_predict_y) // config.OUTPUT_SIZE) - 1, config.NB_PLOT)):
y2.extend(val_y[i * config.OUTPUT_SIZE])
y2b.extend(val_predict_y[i * config.OUTPUT_SIZE])
y3, y3b = [], []
for i in range(
min((len(test_predict_y) // config.OUTPUT_SIZE) - 1, config.NB_PLOT)
):
y3.extend(test_y[i * config.OUTPUT_SIZE])
y3b.extend(test_predict_y[i * config.OUTPUT_SIZE])
y1, y1b, y2, y2b, y3, y3b = (
np.array(y1),
np.array(y1b),
np.array(y2),
np.array(y2b),
np.array(y3),
np.array(y3b),
)
y1 = [y1[i] - y1b[i] for i in range(len(y1))]
y2 = [y2[i] - y2b[i] for i in range(len(y2))]
y3 = [y3[i] - y3b[i] for i in range(len(y3))]
x1 = np.array(list(range(len(y1))))
x2 = np.array(list(range(len(y2)))) + len(x1)
x3 = np.array(list(range(len(y3)))) + len(x1) + len(x2)
plt.plot(x1, y1, label="train", color="green")
plt.plot(x2, y2, label="validation", color="blue")
plt.plot(x3, y3, label="test", color="orange")
# plt.xticks(tick, time, rotation=20)
plt.xlabel("Time")
plt.ylabel("Difference (kW)")
plt.legend()
plt.savefig(outputFolder + "/difference.png")
plt.show()
def plotPart(timestamps, real):
time, tick = makeTick(timestamps, "%H:%M")
x1 = list(range(len(real)))
plt.plot(x1, real, label="house load", color="green")
plt.xticks(tick, time, rotation=20)
plt.xlabel("Time on Day " + timestamps[0].strftime("%m-%d"))
plt.ylabel("Power consumption (kW)")
plt.legend()
plt.tight_layout()
plt.show()
def plotPredictionPart(config, real, predicted, nameOfSet, timestamps, name):
time, tick = makeTick(timestamps, "%H:%M")
x1 = list(range(len(real)))
plt.plot(x1, real, label="actual of " + nameOfSet, color="green")
plt.plot(x1, predicted, label="predicted of " + nameOfSet, color="orange")
plt.xticks(tick, time, rotation=20)
plt.xlabel("Time")
plt.ylabel("Power consumption (kW)")
plt.legend()
plt.tight_layout()
plt.savefig(config.OUTPUT_FOLDER + "/prediction_part-" + name + ".png")
plt.show()
def plotSets(timestamps, train, validation, test):
time, tick = makeTick(timestamps)
x1 = range(len(train))
x2 = range(len(train), len(train) + len(validation))
x3 = range(len(train) + len(validation), len(timestamps))
plt.plot(x1, train, label="train set", color="green")
plt.plot(x2, validation, label="validation set", color="blue")
plt.plot(x3, test, label="test set", color="red")
plt.xticks(tick, time, rotation=20)
plt.xlabel("Time")
plt.ylabel("Power consumption (kW)")
plt.legend()
plt.tight_layout()
plt.show()
def plot_multiple_days(config, loadConfig, test, predicts, timestamps):
predicts = get_following_days(config, predicts)
plt.plot(range(len(test)), test, label="test set")
label = "LSTM lb=" + str(loadConfig.LOOK_BACK)
diff = max(loadConfig.LOOK_BACK, loadConfig.OUTPUT_SIZE) - min(
loadConfig.LOOK_BACK, loadConfig.OUTPUT_SIZE
)
plt.plot(range(loadConfig.LOOK_BACK, len(test) - diff), predicts, label=label)
time, tick = makeTick(timestamps)
plt.xticks(tick, time, rotation=20)
plt.xlabel("Time")
plt.ylabel("Power (kW)")
plt.legend()
plt.tight_layout()
plt.show()
def plotPredictionPartMult(config, real, allPredicted, nameOfSet, timestamps, name):
time, tick = makeTick(timestamps, "%H:%M")
x1 = np.array(list(range(len(real))))
mark = [allPredicted[i][0] for i in range(config.TIME_PER_DAY)]
plt.plot(x1, real, "o-", label="actual " + nameOfSet, color="green")
plt.plot(x1, mark, "x")
for i in range(config.TIME_PER_DAY):
plt.plot(x1 + i, allPredicted[i], linewidth=0.4)
# plt.xticks(tick, time, rotation=20)
plt.xlabel("Time")
plt.ylabel("Power consumption (kW)")
plt.xlim(right=25)
plt.legend()
plt.tight_layout()
plt.savefig(config.OUTPUT_FOLDER + "/prediction_all-" + name + ".png")
plt.show()
def plotInputDay(df, timestamps, config):
realMeans, realSd = getMeanSdDay(config, df)
x1 = list(range(96))
plt.plot(x1, realMeans, label="mean", color="green")
plt.fill_between(
x1, realMeans - realSd * 0.5, realMeans + realSd * 0.5, color="green", alpha=0.5
)
time, tick = makeTick(timestamps[:96], "%H:%M")
plt.xticks(tick, time, rotation=20)
plt.xlabel("Time of Day")
plt.ylabel("Average Power consumption (kW)")
plt.legend()
plt.tight_layout()
plt.show()
def plot_baselines(config, train, test, timestamps):
plt.plot(range(len(test)), test, label="test set")
plt.plot(range(len(test)),
predictMean(config, train, test),
label="mean prediction")
plt.plot(range(1, len(test)),
test[0:-1],
label="next value persistence model")
time, tick = makeTick(timestamps)
plt.xticks(tick, time, rotation=20)
plt.xlabel("Time")
plt.ylabel("Power (kW)")
plt.legend()
plt.tight_layout()
plt.savefig(config.OUTPUT_FOLDER + "/plot_baselines.png")
plt.show()
def plotDayBaseline(config, timestamps, realY, predictY):
timestamps_per_day = get_timestamps_per_day(config)
realMeans, realSd = getMeanSdDayBaseline(config, realY)
x1 = list(range(timestamps_per_day))
plt.plot(x1, realMeans, label="actual", color="green")
plt.fill_between(x1,
realMeans - realSd * 0.5,
realMeans + realSd * 0.5,
color="green",
alpha=0.5)
plt.plot(x1, predictY, label="predict Baseline", color="orange")
time, tick = makeTick(timestamps[:timestamps_per_day], "%H:%M")
plt.xticks(tick, time, rotation=20)
plt.xlabel("Time of Day")
plt.ylabel("Average Power consumption (kW)")
plt.legend()
plt.tight_layout()
plt.savefig(config.OUTPUT_FOLDER + "/plotDayBaseline.png")
plt.show()
def plotPrediction(
train_y,
train_predict_y,
val_y,
val_predict_y,
test_y,
test_predict_y,
timestamps,
config,
):
y1, y1b = [], []
time = []
for i in range(
min((len(train_predict_y) // config.OUTPUT_SIZE) - 1, config.NB_PLOT)
):
y1.extend(train_y[i * config.OUTPUT_SIZE])
y1b.extend(train_predict_y[i * config.OUTPUT_SIZE])
time.append(timestamps[i * config.OUTPUT_SIZE])
y2, y2b = [], []
for i in range(min((len(val_predict_y) // config.OUTPUT_SIZE) - 1, config.NB_PLOT)):
y2.extend(val_y[i * config.OUTPUT_SIZE])
y2b.extend(val_predict_y[i * config.OUTPUT_SIZE])
time.append(timestamps[len(train_predict_y) + i * config.OUTPUT_SIZE])
y3, y3b = [], []
for i in range(
min((len(test_predict_y) // config.OUTPUT_SIZE) - 1, config.NB_PLOT)
):
y3.extend(test_y[i * config.OUTPUT_SIZE])
y3b.extend(test_predict_y[i * config.OUTPUT_SIZE])
time.append(
timestamps[
len(train_predict_y) + len(val_predict_y) + i * config.OUTPUT_SIZE
]
)
y1, y1b, y2, y2b, y3, y3b = (
np.array(y1),
np.array(y1b),
np.array(y2),
np.array(y2b),
np.array(y3),
np.array(y3b),
)
x1 = np.array(list(range(len(y1))))
x2 = np.array(list(range(len(y2)))) + len(x1)
x3 = np.array(list(range(len(y3)))) + len(x1) + len(x2)
time, tick = makeTick(time, "%H:%M")
tick = [i * config.TIME_PER_DAY for i in tick]
plt.plot(x1, y1, label="actual - train", color="green")
plt.plot(x1, y1b, label="predict - train", color="orange")
plt.plot(x2, y2, label="actual - val", color="blue")
plt.plot(x2, y2b, label="predict - val", color="red")
plt.plot(x3, y3, label="actual - test", color="green")
plt.plot(x3, y3b, label="predict - test", color="orange")
plt.xticks(tick, time, rotation=20)
plt.xlabel("Time")
plt.ylabel("Power output (kW)")
plt.legend()
plt.savefig(config.OUTPUT_FOLDER + "/prediction.png")
plt.show()