Website Traffic Forecasting means forecasting traffic on a website during a particular period. It is one of the best use cases of Time Series Forecasting. If you want to learn how to forecast traffic on a website, this article is for you. In this article, I will take you through the task of Website Traffic Forecasting using Python.
The dataset I am using for Website Traffic Forecasting is collected from the daily traffic data of thecleverprogrammer.com. It contains data about daily traffic data from June 2021 to June 2022. You can download the dataset from here. Now let’s get started with the task of website traffic forecasting by importing the necessary Python libraries and the dataset: 1
import pandas as pd
import matplotlib.pyplot as plt
import plotly.express as px
import plotly.graph_objects as go
from statsmodels.tsa.seasonal import seasonal_decompose
from statsmodels.graphics.tsaplots import plot_pacf
from statsmodels.tsa.arima_model import ARIMA
import statsmodels.api as sm
data = pd.read_csv("Thecleverprogrammer.csv")print(data.head())
Date Views 0 01/06/2021 7831 1 02/06/2021 7798 2 03/06/2021 7401 3 04/06/2021 7054 4 05/06/2021 7973
The dataset contains two columns, date and traffic. Before moving forward, I will convert the Date column into Datetime data type: 1
data["Date"] = pd.to_datetime(data["Date"], format="%d/%m/%Y")
print(data.info())
<class 'pandas.core.frame.DataFrame'> RangeIndex: 391 entries, 0 to 390 Data columns (total 2 columns): # Column Non-Null Count Dtype --- ------ -------------- ----- 0 Date 391 non-null datetime64[ns] 1 Views 391 non-null int64 dtypes: datetime64[ns](1), int64(1) memory usage: 6.2 KB None
The Date time column was an object initially, so I converted it into a Datetime column. Now let’s have a look at the daily traffic of the website: 1
plt.style.use('fivethirtyeight')plt.figure(figsize=(15, 10))
plt.plot(data["Date"], data["Views"])
plt.title("Daily Traffic of Thecleverprogrammer.com")plt.show()
Our website traffic data is seasonal because the traffic on the website increases during the weekdays and decreases during the weekends. It is valuable to know if the dataset is seasonal or not while working on the problem of Time Series Forecasting. Below is how we can have a look at whether our dataset is stationary or seasonal: 1
result = seasonal_decompose(data["Views"], model='multiplicative', freq = 30)
fig = plt.figure()
fig = result.plot()
fig.set_size_inches(15, 10)
I will be using the Seasonal ARIMA (SARIMA) model to forecast traffic on the website. Before using the SARIMA model, it is necessary to find the p, d, and q values. You can learn how to find p, d, and q values from here. As the data is not stationary, the value of d is 1. To find the values of p and q, we can use the autocorrelation and partial autocorrelation plots: 1
pd.plotting.autocorrelation_plot(data["Views"])
plot_pacf(data["Views"], lags = 100)
Now here’s how we can train a SARIMA model for the task of website traffic forecasting: 1
p, d, q = 5, 1, 2
model=sm.tsa.statespace.SARIMAX(data['Views'], order=(p, d, q), seasonal_order=(p, d, q, 12))
model=model.fit()
print(model.summary())
Statespace Model Results ========================================================================================== Dep. Variable: Views No. Observations: 391 Model: SARIMAX(5, 1, 2)x(5, 1, 2, 12) Log Likelihood -3099.402 Date: Tue, 28 Jun 2022 AIC 6228.803 Time: 07:01:10 BIC 6287.827 Sample: 0 HQIC 6252.229 - 391 Covariance Type: opg ============================================================================== coef std err z P>|z| [0.025 0.975] ------------------------------------------------------------------------------ ar.L1 0.7808 0.134 5.836 0.000 0.519 1.043 ar.L2 -0.7973 0.135 -5.920 0.000 -1.061 -0.533 ar.L3 -0.1442 0.170 -0.850 0.395 -0.477 0.188 ar.L4 -0.1833 0.151 -1.210 0.226 -0.480 0.114 ar.L5 -0.1548 0.139 -1.117 0.264 -0.426 0.117 ma.L1 -1.1826 0.094 -12.515 0.000 -1.368 -0.997 ma.L2 0.8856 0.078 11.304 0.000 0.732 1.039 ar.S.L12 -0.2606 4.608 -0.057 0.955 -9.293 8.772 ar.S.L24 0.0428 0.781 0.055 0.956 -1.488 1.573 ar.S.L36 -0.1880 0.246 -0.764 0.445 -0.670 0.294 ar.S.L48 -0.2151 0.959 -0.224 0.823 -2.095 1.664 ar.S.L60 0.0127 0.986 0.013 0.990 -1.920 1.946 ma.S.L12 -0.6902 4.611 -0.150 0.881 -9.728 8.348 ma.S.L24 -0.0994 3.637 -0.027 0.978 -7.228 7.029 sigma2 1.257e+06 1.59e+05 7.914 0.000 9.46e+05 1.57e+06 =================================================================================== Ljung-Box (Q): 102.98 Jarque-Bera (JB): 1.32 Prob(Q): 0.00 Prob(JB): 0.52 Heteroskedasticity (H): 1.03 Skew: 0.14 Prob(H) (two-sided): 0.85 Kurtosis: 3.01 ===================================================================================
Now let’s forecast traffic on the website for the next 50 days: 1
predictions = model.predict(len(data), len(data)+50)
print(predictions)
391 9874.390136 392 10786.957398 393 10757.445305 394 9863.890552 395 8765.031698 396 8212.310651 397 8929.181869 398 9685.809771 399 10270.622236 400 10625.904093 401 9854.870630 402 9362.193417 403 9040.021193 404 9081.558484 405 10538.993124 406 11003.816870 407 10897.859601 408 10083.291284 409 9445.806523 410 8629.901288 411 9184.420361 412 10392.770399 413 10593.941868 414 10788.128238 415 10263.101427 416 9449.467789 417 9040.226113 418 9168.972091 419 9887.094079 420 10218.658067 421 10715.657122 422 9899.224399 423 9541.622897 424 9065.810941 425 8825.335634 426 10137.936392 427 10839.866240 428 10905.862922 429 10411.640309 430 9451.211368 431 8698.339931 432 8725.534103 433 10060.678587 434 10506.263524 435 10842.515622 436 10485.387495 437 9335.244813 438 9175.122336 439 9357.034382 440 10295.910655 441 11162.934817 dtype: float64
Here’s how we can plot the predictions: 1
data["Views"].plot(legend=True, label="Training Data", figsize=(15, 10))
predictions.plot(legend=True, label="Predictions")
So this is how you can forecast website traffic for a particular period. Website traffic prediction is one of the best data science project ideas you can mention on your resume. I hope this article has been helpful for you to learn website traffic prediction using the Python programming language. Feel free to ask valuable questions in the comments section below.
