Matplotlib time series historical vs prediction plot shifting one month in the x-axis-CodePudding

I'm trying to plot historical and predicted data into a single line in a single plot. The data ranges from '2020-01-01' '2021-12-31'. When the plot is produced, it plots the data from one month shifted to right in the x-axis. My code is given as follows.

import pandas as pd

index = pd.date_range(start='2020-01-01', end='2021-12-31', freq='M')
series = pd.DataFrame({'num':np.random.randint(0, 50, size=(len(index)))}, index=index)

def plot_series(series):
    import matplotlib.pyplot as plt
    import matplotlib.dates as mdates
    plt.style.use('ggplot')
    
    fig = plt.figure(figsize=(18, 6))
    ax = plt.axes()
    plt.plot(series.index, series.num.values, color='black', label='Historical')
    plt.plot(series.index[18:], series.num.values[18:], color='red', label='Forecasted')
    ax.yaxis.set_ticks(np.arange(0, 105, 5))
    ax.xaxis.set_major_locator(mdates.MonthLocator(interval=1))
    ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m-%d'))
    plt.legend(loc='best')
    plt.title("Time Series Plot: Historial to Forecasted")
    plt.xlabel("Month")
    plt.ylabel("Score");
    plt.gcf().autofmt_xdate(rotation=80)
    plt.show();
    return fig

Here is the resulting plot.

Why this behavior is happening and how can fix it?

CodePudding user response：

If you inspect/print the index object that you generate, you'll see that the dates are at month's end. The

CodePudding user response：

See the inline comments
- ax.set_xticks(ax.get_xticks()[1:-1]): set the x-ticks to remove the undesired ticks at the beginning and end of the axis.
- ax.margins(0.015, tight=True): or set the margins to be small and tight
The code was adjusted to only use the object oriented approach, axes.
Don't include imports inside the function, they should all be at the top of the script.
Tested in python 3.8.11, matplotlib 3.4.3

import matplotlib.pyplot as plt
import matplotlib.dates as mdates
import numpy as np
import pandas as pd

index = pd.date_range(start='2020-01-01', end='2021-12-31', freq='M')
np.random.seed(365)  # makes the sample data the same each time
series = pd.DataFrame({'num':np.random.randint(0, 50, size=(len(index)))}, index=index)

def plot_series(series):
    fig, ax = plt.subplots(figsize=(18, 6))
    ax.plot(series.index, series.num.values, color='black', label='Historical')
    ax.plot(series.index[18:], series.num.values[18:], color='red', label='Forecasted')
    ax.yaxis.set_ticks(np.arange(0, 105, 5))
    ax.xaxis.set_major_locator(mdates.MonthLocator(interval=1))
    ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m-%d'))
    
#     ax.set_xticks(ax.get_xticks()[1:-1])  # set the x-ticks to remove the first and list value

    ax.legend(loc='best')
    ax.set(title='Time Series Plot: Historial to Forecasted', xlabel='Month', ylabel='Score')
    
    ax.margins(0.015, tight=True)  # or set the margins to be small and tight
    
    fig.autofmt_xdate(rotation=80)
    fig.tight_layout()


plot_series(series)

CodePudding user response：

One way I found to fix the x-axis is to use the shift() function. I shifted the index to one step left: series.index.shift(-1)

Here is the final function:

def plot_series(series):
    import matplotlib.pyplot as plt
    import matplotlib.dates as mdates
    plt.style.use('ggplot')
    
    fig = plt.figure(figsize=(18, 6))
    ax = plt.axes()
    plt.plot(series.index.shift(-1), series.num.values, color='black', label='Historical')
    plt.plot(series.index.shift(-1)[18:], series.num.values[18:], color='red', label='Forecasted')
    ax.yaxis.set_ticks(np.arange(0, 105, 5))
    ax.xaxis.set_major_locator(mdates.MonthLocator(interval=1))
    ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m-%d'))
    plt.legend(loc='best')
    plt.title("Time Series Plot: Historial to Forecasted")
    plt.xlabel("Month")
    plt.ylabel("Score");
    plt.gcf().autofmt_xdate(rotation=80)
    plt.show();
    return fig