Testing `pmdarima` for auto ARIMA¶

Playing around with the pmdarima library for basic ARIMA and ARIMAX models in python - a nice alternative to statsmodels. Time series modeling in Python in general seems very scattered compared to the experience in R unfortunately.

From pmdarima toy datasets: https://alkaline-ml.com/pmdarima/modules/datasets.html

In [1]:

import numpy as np
import pandas as pd
import pmdarima as pm
from pmdarima import model_selection

import matplotlib.pyplot as plt
import plotly.express as px
import plotly.graph_objects as go

In [2]:

# Load a local helper file with the data cleaning and all that
from create_dataset import load_combined_dataset

Using the sample datasets from the pmdarima library we can create a combined dataframe with data from 1980 to 1994 covering Australian beer production (in megaliters), residents, and wine sales (in bottles)

In [3]:

df = load_combined_dataset()
df.head()

Out[3]:

	timeperiod	Wine	Residents	Beer
0	19801	51885.0	14515.7	513.0
1	19802	54954.0	14554.9	427.0
2	19803	67765.0	14602.5	473.0
3	19804	79117.0	14646.4	526.0
4	19811	53013.0	14695.4	548.0

In [4]:

fig = go.Figure()

fig.add_trace(go.Scatter(x=df.index, y=df.Wine, name="Wine Sales (bottles)"))
fig.add_trace(go.Scatter(x=df.index, y=df.Residents, name="Residents"))
fig.update_layout(title="Wine Sales and Residents Over Time", 
                  xaxis_title="Quarters since Q1 1980", yaxis_title="Quarterly Total")

Testing pmdarima for auto ARIMA¶

Testing `pmdarima` for auto ARIMA¶