Skip to content
All projects
2024

Educational Analytics: OULAD Click-Stream Forecasting

Forecasting daily student activity on the Open University learning platform with SARIMA, ARIMAX, Prophet and a custom 1D CNN. The CNN reached MAE 0.20 on log-scale clicks on a chronological hold-out.

400+ MB

Source data

808

Days in seasonal model

4

Models tested

0.20

CNN hold-out MAE (log)

Hypothesis

How well can classical time-series models and a small 1D CNN forecast daily student activity on the Open University learning platform?

Problem
Use of an online learning platform swings sharply around term starts, deadlines and holidays. Forecasting daily activity helps plan server capacity and spot unusual days.
My role
Team of 3 (Hardware and Software for Big Data, Module B, Federico II) with Farshad Farahtaj and Zahra Jafarinejad; I built the pipeline.
Result
A 1D CNN on 30-day windows reached a mean absolute error of 0.20 on log daily clicks over the last 20% of the series.

Honest limits

  • Only the CNN has a real hold-out score; the other models were fitted or scored on their own training data, so they are not ranked.
  • No simple baseline (for example "same as 7 days ago") was run, so 0.20 cannot be called good or bad on its own.
  • Calendar dates are assumed (February and October course starts), and errors are on the log scale.
Overview

Forecasting student VLE engagement on the Open University dataset

The Open University Learning Analytics dataset (OULAD) is a public release (over 400 MB) of click-stream and assessment data across multiple UK course presentations. I aggregated it to about 2.3 years of daily totals of VLE interactions (808 days used by the seasonal model), log-transformed, and built four forecasters: SARIMA, ARIMAX, Prophet, and a custom 1D CNN.

University team project (Hardware and Software for Big Data, Federico II, July 2024) with Farshad Farahtaj and Zahra Jafarinejad; I built the pipeline.

Feature engineering

Two different lag windows for two different model families

Statistical models (SARIMA, ARIMAX, Prophet) and the CNN want different feature shapes, so I built two parallel windows:

  • 7 lag features at 1, 2, 3, 7, 14, 21, and 28 days for the seasonal ARIMA models (Prophet uses no lags), mirroring weekly and monthly seasonality.
  • 30-lag input window for the 1D CNN, letting the network learn its own short-horizon shape without having to write seasonality terms by hand.
Results

What was measured, and what was not

The 1D CNN (TensorFlow/Keras, 50 epochs, batch 32, 30-day input window) reached an MAE of 0.20 on log-scale daily clicks on a chronological hold-out.

SARIMA, ARIMAX and Prophet were fitted and plotted, but they were not scored on the same hold-out as the CNN, so the numbers are not a like-for-like comparison and no model is declared the winner. Errors are on the log scale; they are not percentage errors on raw click counts.

Forecasts

What each model sees and predicts

Three notebook outputs: the raw time series with anomalies flagged, Prophet's forecast against observed clicks in the original scale, and the CNN's training-data fit plus its out-of-sample predictions.

Notebook output
Log clicks over time with anomalies highlighted
Daily VLE log clicks across about 2.3 years. Red dots are statistical anomalies.How to read it: blue is daily clicks on log scale, red dots are points that fall outside the statistical band. Spikes line up with the assumed course starts (February and October); dips in summer and at Christmas. Lags up to 28 days matter.
Notebook output
Prophet forecast on original scale showing observed vs forecast
Prophet forecast versus observed clicks in the original scale.How to read it: blue is the observed clicks, orange is Prophet's forecast. Prophet captures the weekly cycle and the multi-year trend but under-shoots the sharp peaks.
Notebook output
Custom 1D CNN forecast vs observed clicks
Custom 1D CNN prediction versus training data.How to read it: blue is training data, orange is the CNN's forecast on the chronological hold-out (MAE 0.20 on log clicks); dates are shifted by the 30-day window and actual values are not plotted. The CNN follows the shape without explicit seasonality terms. The other models were not scored on the same hold-out, so the CNN is not ranked against them.
Tech stack

Frameworks and infrastructure

PythonTensorFlow/KerasSARIMAARIMAXProphetPandasStreamlit

Source code on GitHub.