Revue d'économie et de statistique appliquée
Volume 22, Numéro 2, Pages 114-130
2025-12-09

Short-term Forecasting Of Household Waste Generation In Algiers: Statistical And Deep Learning Approaches



Authors : Brahimi Nadjib Allah . Taoussi Brahim .

Abstract

Algiers, the largest city in Algeria, is home to over 4 million inhabitants and serves as the country’s vibrant economic hub. As one of the most densely populated cities, it faces significant challenges in managing household waste efficiently. Accurate forecasting of household waste generation is essential to address this issue, as it facilitates better resource allocation, optimizes waste collection schedules, and ensures adequate infrastructure to handle increasing waste volumes. Additionally, accurate forecasts inform decision-making for the planning and development of recycling facilities and landfills. They also quantify environmental risks, such as pollution and greenhouse gas emissions, thereby promoting sustainable waste management practices. Statistical time series techniques, such as Seasonal Autoregressive Integrated Moving Averages (SARIMA), are well-suited for short-term predictions due to their simplicity, interpretability, and minimal data requirements. However, these methods may fail to capture complex nonlinear dependencies in waste generation patterns. In contrast, deep learning models like Recurrent Neural Networks (RNN) and Long Short-Term Memory (LSTM) networks excel at identifying intricate patterns in data. Despite their proficiency, RNN and LSTM models demand large datasets, careful hyperparameter tuning, and are less interpretable. In this study, we analyzed daily household waste weights recorded in Algiers between January 2022 and October 2024. Descriptive analysis revealed that waste generation is positively skewed, with a mean of 2112 tons and a standard deviation of 266 tons. Unusually high values were observed during the annual Eid al-Adha festivities, while Fridays consistently showed lower waste volumes compared to other weekdays. Using the 80/20 split for training and testing, we evaluated model performance with four goodness-of-fit metrics. Our findings demonstrated that the SARIMA model, with a seasonal component of seven days, outperformed the LSTM and RNN models in predicting daily household waste generation in Algiers. Future work could explore the integration of hybrid models that combine the strengths of statistical and deep learning techniques to improve accuracy and address the limitations of individual approaches.

Keywords

Household waste generation ; Short-term forecasting ; Seasonal ARIMA ; Deep learning models ; Algeria

Les articles similaires

أهمية تدوير النفايات المنزلية العضوية والسائلة في تموين القطاع الفلاحي (دراسة حالة مؤسسة (epwg/cet) للنفايات المنزلية بولاية عين تموشنت). The Importance Of Recycling Organic And Liquid Household Waste In Supplying The Agricultural Sector. (a Case Study Of (epwg/cet) Enterprise For Household Waste In The Wilaya Of Ain Temouchent)

درويش عمار . 
ص 772-792.

Predictive Modeling Of Ozone Dosing In Drinking Water Treatment Plant Using Deep Learning Comparative Study Between Deep Neural Networks And Convolutional Neural Networks

Hellal Aouatef .  Djeddou Messaoud .  Loukam Imed .  A. Hameed Ibrahim .  Al Dallal Jehad .  Shawaqfah Moayyad . 
pages 69-83.

Electricity Consumption Forecasting In Algeria Using Arima And Long Short-term Memory Neural Network

Sahed Abdelkader .  Kahoui Hacene . 
pages 78-88.

Comparative Analysis Of Gradient Boosting Machines And Long Short-term Memory Networks For Stream Flow Forecasting

Shaikh Ayub .  More Satish .  Bhirud Yuvaraj .  Shelar Vaibhav .  Bagwan Khalid . 
pages 7-32.

The Effect Of Interval Training (short-short, Short) On Rsa Speed Repetition Efficiency In Under 19-year-old Football Players.

Hawamdeh Younes .  Yahiaoui Mohamed . 
pages 92-108.

Formats de citation