Pattern Analysis and Forecasting of daily traffic of a university website Based on Trend and Seasonal Regression

Project info

• Field of studyInformatics and Information Systems
• LevelMaster's thesis (S2)
• Method typeTime Series
• Estimated build time± 5 working days
• Toolsscikit-learn LinearRegression, numpy, metrics
• PriceUS$25 - US$147

Research goal

Forecast daily visits with search-engine visit ratio and weekday index using Trend and Seasonal Regression, then compare forecast errors on a hold-out test set.

Variables

Example result

• 0.791R2 (in-sample)
• 2.508RMSE (holdout)
• 5.193MAPE % (holdout)
• 0.132Trend slope / month
• 5.374Seasonal amplitude
• 2.8Residual SD

The trend + seasonal regression model forecasts Daily visits 12 periods ahead with RMSE 2.508 and MAPE 5.19%. The trend is 0.132 per month with a seasonal amplitude of 5.37.

Computed from illustrative synthetic data, not from a real study.

About Trend and Seasonal Regression

Trend and seasonal regression regresses a time series on a time index and seasonal dummy variables. It suits data with a clear trend and seasonal pattern where interpreting the components matters more than maximum accuracy.

Steps

1. 1Build a time index t and month dummies (drop one)
2. 2Split the data chronologically
3. 3Fit linear regression on t and the dummies
4. 4Forecast the test period
5. 5Compute MAE, RMSE, and MAPE
6. 6Plot data, trend component, and forecast

Assumptions to check

• The trend can be approximated as linear (or polynomial)
• The seasonal effect is constant across years
• Residuals are not strongly autocorrelated

Reading and reporting the results

Interpret the trend coefficient as the increase per period and the dummy coefficients as seasonal differences from the reference month. Report forecast errors and check residual autocorrelation.

Example Python code

A short starting script with this project's variable names. The delivered version is adapted to your data and fully commented.

import numpy as np
import matplotlib.pyplot as plt
from sklearn.linear_model import LinearRegression

rng = np.random.default_rng(13)
n = 60
search_engine_visit_ratio = np.arange(n)
month = search_engine_visit_ratio % 12
daily_visits = 200 + 2.5 * search_engine_visit_ratio + 15 * np.sin(2 * np.pi * month / 12) + rng.normal(0, 4, n)

dummies = np.eye(12)[month][:, 1:]
X = np.column_stack([search_engine_visit_ratio, dummies])
X_tr, X_te, y_tr, y_te = X[:48], X[48:], daily_visits[:48], daily_visits[48:]

model = LinearRegression().fit(X_tr, y_tr)
pred = model.predict(X_te)
mae = np.abs(y_te - pred).mean()
rmse = np.sqrt(((y_te - pred) ** 2).mean())
mape = (np.abs(y_te - pred) / np.abs(y_te)).mean() * 100
print("Trend per period:", round(model.coef_[0], 3))
print("MAE:", round(mae, 2), "RMSE:", round(rmse, 2), "MAPE %:", round(mape, 2))

plt.plot(search_engine_visit_ratio, daily_visits, label="Data")
plt.plot(search_engine_visit_ratio[:48], model.predict(X_tr), label="Fitted")
plt.plot(search_engine_visit_ratio[48:], pred, "r", label="Forecast")
plt.legend()
plt.show()

Packages and prices

Choose how much help you need. Each higher package includes everything in the lower ones. Prices exclude 11% VAT.

• Payment: 50% down payment before work starts, balance after the deliverables are handed over
• Tax: Prices exclude 11% VAT
• Data and academic integrity: Results are computed from your own research data; you write and stand behind your manuscript, we help with the analysis and your understanding of it
• Revisions: Up to 2 revisions are included in the price (Package 3: 3 rounds)

Frequently asked questions