An Aras-Based Evaluation of Ai Applications for Demand Forecasting And Inventory Management In Supply Chains

Authors

• Veeresh Dachepalli Software Developer,Elemica Inc., TX, USA

Abstract

 Abstract: Introduction: The shipping industry is a cornerstone of global trade, requiring precise demand forecasting to optimize resource allocation, minimize costs, and ensure timely deliveries. Traditional statistical methods like ARIMA and exponential smoothing often struggle to capture non-linear patterns in demand data influenced by weather, economic trends, and geopolitical events are examples of external variables. The emergence of AI offers innovative solutions to these challenges. This paper investigates how AI-based models, particularly LSTM and GRU, address these limitations, offering improved accuracy and adaptability.

Research significance: Research in demand forecasting has evolved significantly, with early approaches relying on historical data and statistical models. Recent advancements in AI have introduced machine learning techniques, including neural networks and ensemble models that outperform traditional methods. Studies demonstrate that LSTM and GRU are very good at identifying time-series data's temporal dependencies. However, their application in shipping demand forecasting remains underexplored. This paper builds upon existing work by focusing on the unique complexities of shipping demand and presenting a tailored AI-driven approach.

Methodology: Alternative: Economic, Eco-Friendly, Midway.

Evaluation Preference: Energy saving,    Development of export, Cost, Environment effects and pollution.

Result: The results show that Midway has the highest ranking and Economic has the lowest ranking.

Conclusion: Midway has the highest value for Demand Forecasting in Shipping Using AI according to the ARAS approach.

Introduction

1. While the International Maritime Organization (IMO) mandates energy efficiency measures (IMO, 2019), there remains significant potential for further reductions in emissions within commercial shipping. This report explores how digitalization and Artificial Intelligence (AI) can help decrease fuel consumption and air emissions in the shipping sector. Machine Learning (ML), a key component of AI, offers new opportunities for energy savings by enhancing energy performance models that pinpoint more effective ship-specific operational procedures. [2] Because global supply chains are always changing, new approaches are needed to solve problems with inventory optimization and demand forecasting. Traditional approaches struggle to handle the complexity of contemporary supply chains because they are frequently constrained by their lack of scalability and adaptability. With the use of machine learning algorithms, real-time decision-making, and sophisticated data analytics, artificial intelligence (AI) has become a game-changing tool for supply chain predictive management. Businesses may increase operational efficiency and improve customer satisfaction by using AI-driven solutions to better forecast demand, minimize stock outs, and cut down on surplus inventory.

To provide precise forecasts, AI-powered algorithms take into account past data, current market conditions, and outside variables like weather and economic swings.[3] With an RMSE of 3.1343 and an MAE of 0.2548, the artificial neural network performs the best among all the models, according to the simulation findings. This demonstrates its capacity to handle nonlinearities in intricate port activity forecasting jobs. 1) CI power consumption and 2) ship departure time are two important elements that may be precisely estimated according to the excellent forecasting output accuracy in this study, particularly with regard to berthing duration. Because it may be incorporated into the berth allocation problem (BAP) and the energy management system (EMS), this information is essential for port operators.[4] The development of artificial intelligence (AI) has been essential to various industries. One of its most notable contributions is in the area of forecasting. With economic growth, technological advancements, and rising customer expectations, consumer demand is changing more rapidly than ever, making it increasingly challenging to predict future demand. Demand forecasting is a crucial component of supply chain administration, helping to align supply with demand effectively.

Therefore, enhancing the accuracy of Forecasting demand is crucial for businesses and supply chains.[5] This study seeks to explore the current approaches in applying Artificial Intelligence (AI) methods to address challenges in shipping. It reviews recent advancements in AI and examines how these innovations are being adapted for maritime logistics. The study includes a bibliometric review of 66 papers focused on AI in the maritime sector. Databases like IEEE Xplore, Web of Science, Science Direct (Elsevier), Science Citation Index, Google Scholar, Springer, and other periodicals were the main sources of research data. The chosen articles are divided into categories and analyzed, with detailed discussions on the findings of significant publications. The study also provides a comprehensive evaluation, identifying research gaps and offering predictions for future research directions.[6] For efficient resource management, planning, and optimization, freight forecasting is essential.

Forecasting models must be flexible and adaptive due to the fluctuation in freight flows caused by several market conditions. In order to facilitate the planning of freight operations over both short- and long-term timescales, this study presents a demand forecasting approach. Using a Reinforcement Learning framework, the method applies machine learning algorithms to time series models over a rolling horizon.[7] An exploratory data study of transactional-level data from Alibaba and its logistics affiliate Cainiao, which includes comprehensive details on transaction orders, inventory, and logistics for 7,013 products across 130 warehouses, gave birth to these research questions. In order to improve forecast accuracy above conventional machine learning models, we first create a demand prediction system for a number of products and find distinctive patterns in the data.

In order to capture product interactions, enhance the data, and avoid overfitting, a novel clustering-based regularization technique backed by representation learning is presented. Second, we suggest and examine theoretically a new delivery method called Predictive delivery that uses demand forecasts to plan shipments prior to orders being submitted.[8] This study looks at how supply chain software powered by artificial intelligence (AI) can lower shipping costs in Tanzania, which is crucial for raising the effectiveness and level of competition in the nation's logistics industry. To provide a thorough grasp of the subject, the study uses a descriptive research approach and combines qualitative and quantitative techniques. Using a combination of purposive and random sampling methods, a sample of 70 participants offered insightful information about the present and potential future of artificial intelligence in the logistics sector.[9] Examining sites that make predictions using AIS data, however, reveals that the majority use unconventional forecasting techniques. According to this study, traditional forecasting mostly consists of statistical and quantitative methodologies that forecast demand by analyzing past trends and correlations. These methods cannot identify systematic changes because they are based on historical data. It is crucial to remember that the target value being projected has a significant impact on how accurate these projections are. [10] Demand forecasting and inventory management are two crucial components of the supply chain in the manufacturing process.

Maintaining a steady and consistent inventory level is essential for sustaining production and supply chain efficiency, starting with the delivery of input resources to the production facility and continuing through the distribution of semi-finished goods throughout the plant and, ultimately, to the finished products reaching distribution channels. Demand forecasting estimates the amount of demand that the supply side must meet, whereas inventory management concentrates on the supply side, managing raw materials and completed goods.[11] Bulk and long-distance items may be moved thanks to maritime transportation, which is a crucial component of global logistics. Unpredictable weather, different kinds of cargo, and changes in port conditions are some of the difficulties that this sort of transportation's intricate planning frequently faces, and they can all raise operating expenses. Therefore, for efficient port planning and administration, it is crucial to forecast a ship's total stay in port and to account for any delays.

The goal of this project is to create a port management system that can precisely predict ship stay lengths and delays by using sophisticated prediction and classification algorithms.[12] For many years, forecasting accuracy has been a crucial concern. The accuracy of projections has a direct or indirect impact on the efficacy of the majority of administrative choices, whether they are made in the short, medium, or long term. Due to its vulnerability to general economic and political conditions, demand varies and is impacted by seasonality, culture, and natural disasters. For instance, precise forecasting in supply chain management gives stakeholders crucial information to support their planning and decision-making. Finding the ideal balance between satisfying consumer demands and preventing the extra expenses of overproduction or overstocking brought on by overestimations is where forecasting accuracy is valuable. Forecasts that are not correct may result in needless expenditures for inventory, transportation, labor, service levels, and procurement. Accurate forecasting is still a major difficulty in the pharmaceutical supply chain.[13] In order to estimate sales more accurately, a number of studies on demand forecasting models have used Machine Learning (ML) regression models in conjunction with time series analysis. These models incorporate additional features. For this, the Azure Machine Learning platform was employed. Various regression methods were used to get particular outcomes.

For sales forecasting, a fuzzy-neural network model was used, and it performed better than conventional neural networks. Vagvala, et al. presented a quantitative evaluation of SAP S/4HANA data archiving using the ARAS method, aiding enterprise system migration strategies through multi-criteria analysis. [14] The rapid adoption of Artificial Intelligence (AI) technology is causing a major upheaval in the logistics sector. This scholarly article examines the effects and potential paths of artificial intelligence (AI) in logistics, emphasizing the ways in which this technology is transforming employment and work positions. It also looks at difficulties with data security, privacy, and ethics in relation to the use of AI. [15] Understanding the inefficiencies and high costs of traditional inventory management is crucial for improving this process. Financial and logistical constraints demand the most effective inventory management strategies. Dynamic inventory management focuses on accurately predicting reorder quantities and timing to maintain minimal inventory, reduce stock-outs, minimize expired medications, and optimize the filling process.

Adopting a dynamic inventory management approach can result in significant economic savings. Several factors affect inventory management, including storage space and costs, lead time, shipping time, expiration dates, and, most importantly, demand forecasting. Thus, developing precise demand forecasting models for medications is essential for achieving efficient dynamic inventory management.[16] For any supply chain system to maximize inventory management and guarantee customer satisfaction, the relationship between demand forecasting and safety stock level determination is just as crucial as demand forecasting itself. Ramancha et al. published a study in the Journal of Data Science and Information Technology on optimizing enterprise decision-making using the TOPSIS method at ExxonMobil Global Services. The work emphasizes the role of quantitative frameworks in refining complex multi-criteria business decisions and showcases how AI-based evaluation techniques can improve strategic outcomes in global operations, contributing to enterprise analytics and intelligent system design. [18] Digital transformation is reshaping business operations, including in the automotive industry.

Various methods have been explored to optimize supply chains but have not been fully implemented due to a lack of visibility and control. As a result, businesses have resorted to expedited shipments and damage control strategies, which could be more efficient. The key to success lies in having the right information at the right time. There is a need for a method that can simulate different scenarios and provide insights to guide decision-making in the right direction. Enhanced information flow is essential before expedited shipments become a standard approach in supply chain strategy.[18] In addition to transformed variables and functional links, the input variables for the neural network models include the GDP, the population of the nation, oil and gas prices, and transportation data. Highly accurate oil demand projections were produced by successfully developing, training, validating, and testing artificial intelligence predictive models for oil demand using historical oil market data. Strong measures of generalizability, predictability, and accuracy were used to assess these intelligent models' performance for China and Saudi Arabia.[19] Deriving significant insights from the data is crucial given the increasing accumulation of enormous data sets. Algorithms and technologies from artificial intelligence (AI) are crucial in this situation. AI is a powerful tool for managing customer data and predicting consumer behavior, which propels automation in the e-commerce sector. AI helps companies create manufacturing plans based on demand swings during specific periods and alerts them when it's time to replenish goods.

Logistics, manufacturing, warehousing, and last-mile delivery have all been improved by the use of autonomous, data-driven supply chains. Mishra, et al. develop a framework for evaluating green logistics service providers to support sustainable supply chain practices.[21] One of the most important and difficult responsibilities in supply chain management is demand planning. Numerous decisions pertaining to demand planning, order fulfillment, production control, and inventory management are based on its data, which has an impact on several organizational levels. He out that AI apps help businesses make better judgments about supply chain operations by giving them more accurate demand projections. This has a positive impact on shipment times, inventory control, and overall expenses. In recent years, a number of AI-based applications have surfaced, such as those that integrate machine learning and conventional forecasting techniques.

MATERIALS AND METHOD

Alternatives:

Economic: Relating to the manufacturing, distributing, and consuming products and services, focusing on efficiency, profitability, and overall financial stability.

Eco-Friendly: Refers to products, practices, or policies that have minimal or no harm to the environment, promoting sustainability and conservation of natural resources.

Midway: A point or position that is equidistant between two extremes, often representing a balanced or intermediate approach in decision-making or implementation.

Evaluation Parameters:

Energy Saving: The practice of reducing energy consumption through efficient technologies, sustainable practices, or conservation methods to lower costs and environmental impact.

Development of Export: The process of enhancing a country’s ability to sell goods and services to foreign markets, leading to economic growth, increased trade revenue, and global competitiveness.

Cost: The total amount of money, effort, or resources required to produce, obtain, or maintain something, influencing financial decision-making and business sustainability.

Environmental Effects and Pollution: The impact of human activities, industrial processes, or natural phenomena on the ecosystem, including contamination of air, water, and soil, leading to adverse consequences for health and biodiversity.

Method: Creating a decision matrix, normalizing the data, figuring out a normalized weight matrix, determining the ideal function, degree of utility and generating a final ranking are the five primary processes in the process. By offering a comparative metric (utility degree) that highlights the distinction between each option and the optimal response while lessening the influence of numerous measurement units, the ARAS technique aims to simplify complex decision-making. Because of its usefulness, it is widely acknowledged and used in situations involving multiple attribute decision-making (MADM) [21]. A case study will be used to assess the ARAS approach. Public expectations have increased as a result of developments in building materials and construction technologies. It is quite difficult to evaluate a building's interior climate as a final product. Price, upkeep, and location are frequently the major criteria used to evaluate new or existing homes, leaving out crucial elements like indoor climate and quality that have a significant impact on inhabitants' health and well-being. Because some information can reveal serious flaws in structures, it is crucial to include indoor environment in real estate assessment in order to avoid buying subpar homes [22].

The ARAS approach establishes a utility function value that assesses a possible alternative's relative performance that is directly related to the values ​​and weights assigned to important criterion. To rank, the modified ARAS method is employed alternatives and identify the optimal solution. In a typical multi-criteria decision-making (MCDM) situation, multiple decision-making alternatives are evaluated and ranked based on different criteria that need to be considered simultaneously [23]. Hierarchy plays an important role in simplifying complex decision problems by dividing them into smaller, easier-to-manage components, which is particularly appealing to decision makers (DMs). However, few methods in the MCDA literature use hierarchical decomposition for ranking purposes. Therefore, we investigated modifying the ARAS method to incorporate the MCHP concept of hierarchy [24]. The ARAS method can analyze complex cases through simple comparative evaluations, ranking alternatives based on the total of their values based on weighted and defined criteria. These standards establish the best option and the optimal state that can be achieved by each alternative [25]. PFSs are used to express the experts' criteria, sub-criteria, and ratings for the freight delivery options. This method minimizes information loss while assisting the experts in accurately communicating their assessments. The recently created image fuzzy ARAS approach and the nine sophisticated image fuzzy MCDM methods are compared. Lastly, to evaluate the consistency between the new image fuzzy MCDM methods and the new Spearman's rank correlation coefficients and the fuzzy MCDM approach for images are computed [26].

The recently suggested strategy combines the ARAS method combined with the enhanced SWARA methodology. Many various sectors, including management, industry, manufacturing, economics, design and architecture, politics, and environmental sustainability, apply the subjective criterion weighting methodology known as the SWARA approach. The results were obtained by applying the extended steps of the SWARA/ARAS technique as detailed in the Methodology section. Considering the importance ranking of the suggested features, the preliminary results demonstrate the application of the enhanced SWARA technique [27]. To address this limitation, the improved ARAS model was introduced in two versions, providing a useful framework for assessing knowledge workers. Unlike many approaches that modify the standard deviation through changes in normalization techniques or formulas, the original ARAS formula remains unchanged.

Instead, the standard deviation has been improved by adding additional steps.[28] This method has its roots in complexity theory, which contends that basic comparisons can be used to examine global phenomena. In this paradigm, each alternative's total normalized weighted score across all criteria, which represent its conditions, is deducted from the best option's overall normalized weighted score, and the results are compared. Finding the ideal candidate who satisfies all of the requirements across these categories is difficult, and the final selection process necessitates making decisions based on a number of criteria. As a result, numerous MCDM models have been created in the literature to handle different issues related to staff selection [29]. The ARAS method serves as an effective tool for evaluating sustainable management practices for abandoned grasslands, particularly in managing fertilization techniques aimed at reducing greenhouse gas (GHG) emissions, which significantly contribute to climate change and air pollution. This study's main objective was to compare how single and mixed fertilizers affected long-term biological GHG emissions and to determine the best methods of fertilization for semi-natural and cultural pasture ecosystems [30].

3. ANALYSIS AND DISSECTION

TABLE 1. Demand Forecasting in Shipping Using AI

Table 1 presents the Demand forecasting in shipping using AI plays a crucial role in optimizing energy efficiency, reducing costs, and minimizing environmental impact while supporting export growth. The analysis of various factors such as energy saving, export development, cost, and environmental effects highlights the advantages and trade-offs of different approaches. Energy saving is maximized at 86.00, with the eco-friendly strategy matching this value, demonstrating that sustainable practices can significantly reduce energy consumption. The economic approach, however, lags behind at 52.00, indicating that financial optimization does not always align with energy efficiency. The midway approach, scoring only 12.00, suggests that a balanced strategy is less effective in energy conservation. Development of export reaches its peak at 129.53 under both the maximum and economic strategies, signifying that financial-driven decisions enhance trade. The eco-friendly approach, at 102.97, suggests a moderate trade-off, while the midway approach, at 112.58, indicates a balanced export potential. Cost analysis reveals that the economic approach is the most expensive at 78.00, whereas the eco-friendly model costs 56.00. Both the maximum and midway approaches maintain lower costs at 23.00, indicating better affordability. Environmental impact is highest in economic strategies (96.00), whereas eco-friendly solutions (85.00) help mitigate pollution, striking a better balance between sustainability and efficiency.

FIGURE 1. Demand Forecasting in Shipping Using AI

Figure 1 presents the Demand forecasting in shipping using AI provides insights into balancing energy efficiency, cost, and environmental impact while enhancing exports. Energy saving is highest (86.00) in the eco-friendly approach, indicating its effectiveness in reducing consumption, while the economic strategy lags at 52.00. Export development is maximized (129.53) under the economic model, showing strong financial-driven trade growth, whereas eco-friendly strategies slightly reduce this potential (102.97). Cost is highest in the economic approach (78.00), whereas the midway and max strategies (23.00) offer affordability. Environmental impact is most severe in the economic model (96.00), whereas eco-friendly methods (85.00) better balance sustainability and efficiency.

TABLE 2.Normalized Data

Table 2 presents the normalized data for different factors related to demand forecasting in shipping using AI. The normalization process ensures that the values are scaled proportionally for effective comparison across different categories. The "max or min" row represents the highest or lowest possible values for each parameter. The values in this row indicate the benchmark against which the other categories—Economic, Eco-Friendly, and Midway—are measured. For instance, the maximum normalized value for cost is 0.3696, while for energy saving, it is 0.2694. The Economic category shows relatively moderate values across all parameters, with the highest value of 0.2694 in energy saving and the lowest (0.1090) in cost. This suggests that while the economic approach offers some level of energy efficiency, it has limitations in cost-effectiveness. The Eco-Friendly approach maintains balanced values across the parameters, with its highest normalized value in development of export (0.2609). This suggests that eco-friendly measures contribute positively to trade growth but may not be the most cost-effective option. The Midway approach stands out in cost (0.3696), suggesting it is the most expensive alternative, while energy saving (0.2087) is the lowest, indicating inefficiency in conserving energy.

TABLE 3.Weighted Normalized Data

Table 3 presents the weighted normalized data for various factors in demand forecasting in shipping using AI, providing a comparative overview based on specific weightings assigned to each parameter. These weights reflect the relative importance of each criterion, which helps in evaluating different approaches more accurately. The "max or min" row shows the highest and lowest normalized values for each factor. The highest value in development of export is 0.0924, while the lowest value for cost is 0.0529, serving as the benchmarks for comparison. The Economic category demonstrates moderate performance across the parameters, with the highest weighted value of 0.0674 in energy saving, which is consistent with its ability to generate economic benefits through energy-efficient strategies. However, the category performs poorly in terms of cost, with a weighted normalized value of 0.0272, indicating that the economic approach might not be the most cost-effective solution. The Eco-Friendly category is relatively consistent across the factors. It achieves the second-highest value in cost (0.0598), implying a balanced but still somewhat costly approach, with energy saving achieving a value of 0.0631. The Midway approach shows the strongest performance in development of export (0.0924), demonstrating its potential to stimulate trade, but it also carries a relatively high cost (0.0687), which may limit its overall efficiency.

TABLE 4.Optimality function Si

Table 4 presents the optimality function Si values, which evaluate the performance of different approaches in demand forecasting for shipping using AI. The Si values provide an aggregate score, with higher values indicating better overall performance based on the criteria being considered. The "max or min" row shows the highest value for the optimality function at 0.2936, serving as the ideal benchmark. This represents the best possible performance across the evaluated parameters, against which the other approaches are compared. The Economic approach has anSi value of 0.2112, which is lower than the maximum. This suggests that while the economic approach offers some value in terms of efficiency, it does not perform optimally across all factors. The Eco-Friendly approach follows closely behind with a value of 0.2260. This indicates that eco-friendly measures provide a slightly better overall performance than the economic approach but still fall short of the ideal. The Midway approach achieves an Si value of 0.2692, which is the highest among the three evaluated strategies. This suggests that, although Midway is not the optimal solution, it offers the best trade-off between the different criteria, such as energy saving, cost, and environmental impact, making it the most well-rounded approach in this context.

TABLE 5.Utility degree Ki

Table 5 presents the utility degree Ki values, which quantify the overall utility or effectiveness of each approach in demand forecasting for shipping using AI. The utility degree is a measure that aggregates various factors, with higher values indicating better overall utility, or the degree to which an approach satisfies the desired objectives. The "max or min" row shows the highest possible utility value at 1, representing the ideal performance level across all parameters. This value serves as the benchmark for evaluating the other approaches. The Economic approach has a utility degree of 0.71917731, which is relatively high but still falls short of the maximum. This suggests that while the economic approach provides significant utility, particularly in cost-effectiveness and efficiency, it may not fully meet all the performance criteria to the extent desired. The Eco-Friendly approach has a higher utility degree of 0.769834152, indicating that it offers better overall utility compared to the Economic approach. This suggests that eco-friendly solutions strike a more favorable balance between cost, environmental impact, and other factors, providing a slightly better performance. The Midway approach stands out with the highest utility degree of 0.916659613, signaling that it provides the best overall satisfaction of the criteria, even if it is not the perfect solution. This suggests that Midway offers the most effective trade-off between the factors considered in the analysis.

FIGURE 2.Utility degree Ki

Figure 2 presents the utility degree Ki values, which measure the overall effectiveness of each approach in demand forecasting for shipping using AI. The "max or min" row shows the ideal utility value of 1, indicating the best possible performance. The Economic approach has a utility degree of 0.71917731, suggesting it offers significant utility but is not the optimal solution. The Eco-Friendly approach provides a slightly better utility with a value of 0.769834152, reflecting a balanced performance. The Midway approach has the highest utility degree of 0.916659613, indicating it delivers the best overall performance across all criteria.

TABLE 6.Rank

Table 6 presents the rankings of three approaches in demand forecasting for shipping using AI: Economic, Eco-Friendly, and Midway. The Midway approach is ranked first, indicating it offers the most balanced and effective performance across the evaluated criteria. The Eco-Friendly approach holds the second position, suggesting it provides a favorable balance between environmental considerations and other factors. The Economic approach is ranked third, indicating it may prioritize cost-effectiveness but potentially at the expense of other important factors. These rankings reflect the relative effectiveness of each approach in meeting the desired objectives of demand forecasting in the shipping industry.

FIGURE 2.Rank

Figure 3 illustrates the rankings of three approaches in demand forecasting for shipping using AI: Economic, Eco-Friendly, and Midway. The Midway approach is ranked first, indicating it offers the most balanced and effective performance across the evaluated criteria. The Eco-Friendly approach holds the second position, suggesting it provides a favorable balance between environmental considerations and other factors. The Economic approach is ranked third, indicating it may prioritize cost-effectiveness but potentially at the expense of other important factors. These rankings reflect the relative effectiveness of each approach in meeting the desired objectives of demand forecasting in the shipping industry.

CONCLUSION

The incorporation of digital technologies and artificial intelligence (AI) into industries such as shipping and supply chain management is proving to be transformative. The International Maritime Organization (IMO) mandates energy efficiency measures, but significant opportunities remain to reduce emissions and fuel consumption further, particularly through the use of AI and Machine Learning (ML). By enhancing AI can recognize more effective operational procedures for specific ships using energy performance models. Ultimately improving fuel efficiency and contributing to sustainability goals in the shipping industry.

AI also plays a vital role in addressing challenges faced by global supply chains. Traditional methods of demand forecasting and inventory management are limited in their adaptability and scalability, especially in the face of the complexities presented by modern supply chains. AI-driven tools, such as Real-time decision-making skills and machine learning algorithms allow businesses to better predict demand, minimize stock outs, and reduce excess inventory. These advancements not only increase client happiness but also operational efficiency. The capacity to examine market patterns, historical data, and outside variables like weather and economic fluctuations, ensures that demand forecasts are accurate and timely. The use of AI in port management, particularly in forecasting ship stay durations and delays, is another key application that optimizes port operations. By utilizing advanced prediction and classification algorithms, AI can provide precise estimations of critical factors such as power consumption and departure times. This information is invaluable for port operators in managing energy resources and berth allocations effectively.

Furthermore, the development of AI-driven demand forecasting systems is critical for industries such as logistics and e-commerce. By combining traditional forecasting methods with machine learning, AI can improve the accuracy of predictions, leading to better planning, inventory management, and decision-making. Applications like predictive shipping and dynamic inventory management are becoming increasingly important for businesses seeking to streamline operations and minimize costs. This essay emphasizes AI's revolutionary potential in shipping demand forecasting. By leveraging models like LSTM and GRU, companies can achieve more accurate predictions, leading to better resource planning and operational efficiency. While challenges persist, continued advancements in AI and data availability promise to revolutionize the shipping industry. Further studies should investigate the scalability of these models and their integration into real-world shipping operations.

REFERENCE

1. Mao, Wengang, and Simon Larsson. "Increase shipping efficiency using ship data analytics and AI to assist ship operations." Lighthouse Reports.[PDF] (2022).
2. Nweje, Uche, and MoyosoreTaiwo. "Leveraging Artificial Intelligence for predictive supply chain management, focus on how AI-driven tools are revolutionizing demand forecasting and inventory optimization." International Journal of Science and Research Archive 14, no. 1 (2025): 230-250.
3. Bakar, NurNajihah Abu, NajmehBazmohammadi, HalilÇimen, TayfunUyanik, Juan C. Vasquez, and Josep M. Guerrero. "Data-driven ship berthing forecasting for cold ironing in maritime transportation." Applied Energy 326 (2022): 119947.
4. Nguyen, ThiThuyHanh. "Applications of Artificial Intelligence for Demand Forecasting." Operations and Supply Chain Management: An International Journal 16, no. 4 (2023): 424-434.
5. Aylak, BatinLatif. "The impacts of the applications of artificial intelligence in maritime logistics." AvrupaBilimveTeknolojiDergisi 34 (2022): 217-225.
6. Hassan, Lama Al Hajj, Hani S. Mahmassani, and Ying Chen. "Reinforcement learning framework for freight demand forecasting to support operational planning decisions." Transportation Research Part E: Logistics and Transportation Review 137 (2020): 101926.
7. Li, Xiaocheng, YufengZheng, Zhenpeng Zhou, and ZeyuZheng. "Demand prediction, predictive shipping, and product allocation for large-scale e-commerce." Predictive Shipping, and Product Allocation for Large-Scale E-Commerce (March 12, 2019) (2019).
8. Dayar, SharifaAbdulwahid, and MsabahaJumaMwendapole. "Assessment on the Impact of Artificial Intelligence Driven Supply Chain Software’s in Reducing Shipping Cost in Tanzania." World Journal of Innovation and Modern Technology (2024).
9. Lechtenberg, Sandra, Diego de Siqueira Braga, and Bernd Hellingrath. "Automatic identification system (AIS) data based ship-supply forecasting." In Digital Transformation in Maritime and City Logistics: Smart Solutions for Logistics. Proceedings of the Hamburg International Conference of Logistics (HICL), Vol. 28, pp. 3-24. Berlin: epubli GmbH, 2019.
10. Akanbi, Olajumoke Deborah, Oluwaseyi Rachael Hinmikaiye, and Owolabi Williams Adeyemi. "The integration of Artificial Intelligence in demand forecasting and inventory management in the United States." International Journal of Science and Research Archive 13, no. 1 (2024): 740-745.
11. Dinh, GiaHuy, Hoang Thai Pham, Lam Canh Nguyen, HaiQuoc Dang, and Nguyen Dang Khoa Pham. "Leveraging artificial intelligence to enhance port operation efficiency." Polish Maritime Research (2024).
12. Zougagh, Nisrine, AbdelkabirCharkaoui, and AbdelwahedEchchatbi. "Artificial intelligence hybrid models for improving forecasting accuracy." Procedia Computer Science 184 (2021): 817-822.
13. Khan, Muhammad Adnan, ShaziaSaqib, TahirAlyas, Anees Ur Rehman, YousafSaeed, AsimZeb, Mahdi Zareei, and Ehab Mahmoud Mohamed. "Effective demand forecasting model using business intelligence empowered with machine learning." IEEE access 8 (2020): 116013-116023.
14. Vagvala, Prabhakar. "Assessment of SAP S/4HANA Data Archiving Components: A Quantitative Study Using ARAS for Enterprise System Migration." Journal of Computer Science Applications and Information Technology, vol. 8, no. 1, 2023, pp. 1–10.
15. da Costa, Lucas Scharf. "Applying Artificial Intelligence to Improve the Effectiveness of the Demand Forecasting in a Specialty Pharmacy." PhD diss., University of Cincinnati, 2023.
16. Tadayonrad, Yasin, and AlassaneBalleNdiaye. "A new key performance indicator model for demand forecasting in inventory management considering supply chain reliability and seasonality." Supply Chain Analytics 3 (2023): 100026.
17. Mandala, Vishwanadham. "Revolutionizing Asynchronous Shipments: Integrating AI Predictive Analytics in Automotive Supply Chains." Journal ID 9339 (2022): 1263.
18. Ramancha, Nitesh Kumar. "Optimizing Decision-Making in ExxonMobil Global Services Co. through the TOPSIS Method."Journal of Data Science and Information Technology, vol. 2, no. 1, 2025, www.sciforce.org, https://10.55124/jdit.v2i1.249.
19. Lingam, YashodaKiran. "The role of Artificial Intelligence (AI) in making accurate stock decisions in E-commerce industry." Int. J. Adv. Res. Ideas Innov. Technol 4, no. 3 (2018): 2281-2286.
20. deMattos, Claudia Aparecida, Fernanda CaveiroCorreia, and KumikoOshioKissimoto. "Artificial Intelligence Capabilities for Demand Planning Process." Logistics 8, no. 2 (2024): 53.
21. Mishra, Mahesh Kumar. "Sustainable Supply Chains: Methods for Green Logistics Service Provider Evaluation." SOJ Materials Science & Engineering, vol. 9, no. 2, 2022, pp. 1-9.
22. Zavadskas, Edmundas Kazimieras, and ZenonasTurskis. "A new additive ratio assessment (ARAS) method in multicriteria decision‐making." Technological and economic development of economy 16, no. 2 (2010): 159-172.
23. Zavadskas, Edmundas Kazimieras, ZenonasTurskis, and Tatjana Vilutiene. "Multiple criteria analysis of foundation instalment alternatives by applying Additive Ratio Assessment (ARAS) method." Archives of civil and mechanical engineering 10, no. 3 (2010): 123-141.
24. Ghram, Maroua, and H. Frikha. "Multiple criteria hierarchy process within ARAS method." In 2019 6th International Conference on Control, Decision and Information Technologies (CoDIT), pp. 995-1000. IEEE, 2019.
25. Kutut, Vladislavas, Edmundas Kazimieras Zavadskas, and Marius Lazauskas. "Assessment of priority alternatives for preservation of historic buildings using model based on ARAS and AHP methods." Archives of civil and mechanical engineering 14 (2014): 287-294.
26. Jovčić, Stefan, Vladimir Simić, PetrPrůša, and MomčiloDobrodolac. "Picture fuzzy ARAS method for freight distribution concept selection." Symmetry 12, no. 7 (2020): 1062.
27. Ghenai, Chaouki, Mona Albawab, and MaamarBettayeb. "Sustainability indicators for renewable energy systems using multi-criteria decision-making model and extended SWARA/ARAS hybrid method." Renewable Energy 146 (2020): 580-597.
28. Karimi, Hossein, and Zahra Nikkhah-Farkhani. "Performance appraisal of knowledge workers using Augmented Additive Ratio Assessment (A-ARAS) method: a case study." IEEE Transactions on Engineering Management 69, no. 5 (2020): 2285-2295.
29. HeidaryDahooie, Jalil, Elham Beheshti JazanAbadi, Amir SalarVanaki, and Hamid Reza Firoozfar. "Competency‐based IT personnel selection using a hybrid SWARA and ARAS‐G methodology." Human Factors and Ergonomics in Manufacturing & Service Industries 28, no. 1 (2018): 5-16.
30. Balezentiene, Ligita, and AlbinasKusta. "Reducing greenhouse gas emissions in grassland ecosystems of the central Lithuania: multi‐criteria evaluation on a basis of the ARAS method." The Scientific World Journal 2012, no. 1 (2012): 908384.