Historical Series Research Articles

Research on E-Commerce Retail Demand Forecasting Based on SARIMA Model and K-means Clustering Algorithm

With the rapid development of e-commerce, precise demand forecasting and efficient inventory management have become essential for the success and profitability of retail businesses. This study focuses on demand forecasting for e-commerce retailers using the Seasonal Autoregressive Integrated Moving Average (SARIMA) model and the K-means clustering algorithm. The research utilizes a dataset containing 1996 time series of sales data from various products, merchants, and warehouses, aiming to predict demand changes for the next 15 days. The study initially evaluates three models—Linear Regression (LR), Autoregressive Integrated Moving Average (ARIMA), and SARIMA—by fitting them to historical sales data to forecast future demand. The SARIMA model is identified as the most effective through rigorous evaluation using 1-mWAPE (mean weighted absolute percentage error) and RMSE (root mean square error) metrics. To enhance homogeneity within demand categories, the K-means clustering method is applied to divide products into four distinct groups, further refining the forecasting process.The paper also addresses the challenge of integrating new sequences into the dataset by leveraging clustering results to classify sequences and using cosine similarity to identify analogous historical time series. These matched sequences serve as the basis for demand prediction using the established SARIMA model. The findings highlight the robustness of the SARIMA model in capturing trends and seasonality, providing a reliable framework for e-commerce demand forecasting that can significantly impact inventory strategies and operational efficiency.

Twitter's capacity to forecast tourism demand: the case of way of Saint James

PurposeRadical changes in consumer habits induced by the coronavirus disease (COVID-19) pandemic suggest that the usual demand forecasting techniques based on historical series are questionable. This is particularly true for hospitality demand, which has been dramatically affected by the pandemic. Accordingly, we investigate the suitability of tourists’ activity on Twitter as a predictor of hospitality demand in the Way of Saint James – an important pilgrimage tourism destination.Design/methodology/approachThis study compares the predictive performance of the seasonal autoregressive integrated moving average (SARIMA) time-series model with that of the SARIMA with an exogenous variables (SARIMAX) model to forecast hotel tourism demand. For this, 110,456 tweets posted on Twitter between January 2018 and September 2022 are used as exogenous variables.FindingsThe results confirm that the predictions of traditional time-series models for tourist demand can be significantly improved by including tourist activity on Twitter. Twitter data could be an effective tool for improving the forecasting accuracy of tourism demand in real-time, which has relevant implications for tourism management. This study also provides a better understanding of tourists’ digital footprints in pilgrimage tourism.Originality/valueThis study contributes to the scarce literature on the digitalisation of pilgrimage tourism and forecasting hotel demand using a new methodological framework based on Twitter user-generated content. This can enable hospitality industry practitioners to convert social media data into relevant information for hospitality management.

The freshwater discharge into the Adriatic Sea revisited

The present study reconstructs the river discharge climatology and its respective historical series for all rivers of the Adriatic Sea with averaged climatological daily river discharge above 1 m3s−1, to reach a better representation of the Adriatic rivers in hydrodynamic models and, consequently, to develop a more realistic freshwater balance in the different regions of the hydrographic basin. Based on the European Flood Awareness System (EFAS) data set, a careful method of identification and selection of the Adriatic rivers, followed by a rigorous assessment against observational data, was developed to evaluate the current state of the Adriatic river discharges and their respective trends throughout several climate indicators from 1991 to 2022. Observational data are limited to 85% of the identified rivers, totaling 98% of the overall freshwater input into the Adriatic Sea. The results confirm that the Shallow Northern Adriatic receives the largest freshwater inputs with a daily average exceeding 2,400 m3s−1, which amounts to 61% of the overall Adriatic discharges. Consequently, this region guides the freshwater seasonal cycle of the Adriatic Sea, which presents a well-defined pattern of two flood peaks in late autumn and late spring, separated by a minimum discharge period at mid-summer. From the Central to the Southern Adriatic subregions, the absence of snow-melting effects prevents the secondary flood peak during the spring, shaping the seasonal cycle of river discharges from a single flood peak in late autumn to a drought period in August. The 32 years of continuous river discharge data reveal a negligible trend in the overall Adriatic Sea but a negative trend for the last decade (2013–2022). This decadal decrease is driven by the extreme drought that drastically pounded the northern Adriatic in 2022.

Strong coupling between soil moisture and temperature intensifies warming and humidification on the Tibetan Plateau: Evidence from 200-year tree ring records

In recent times, the warming and humidification of the Tibetan Plateau have become more pronounced. However, research on soil moisture on the Tibetan Plateau has been limited by the late initiation of data acquisition, thereby constraining the study of its driving mechanisms. Hence, there is an urgent need for historical time series to fill this research gap. To address this issue, we analyzed four spruce (Picea likiangensis) chronologies collected in high-altitude regions of the Tibetan Plateau, and we found that soil moisture during the dry season was a critical limiting factor for tree growth. The first principal component of a principal component analysis, accounting for 56.21% of the variance in all chronologies, was then used to reconstruct soil moisture variations from October of the previous year to May of the current year in the southeastern Tibetan Plateau from 1778 to 2020. We found significant relationships between soil moisture changes and the El Niño-Southern Oscillation (ENSO), Pacific Decadal Oscillation (PDO), and Atlantic Multidecadal Variability (AMV). We also found a strong linkage between dry season soil moisture and summer temperature, suggesting that this relationship contributes to the intensification of warming and humidification on the Tibetan Plateau.

ClimateDT: A Global Scale-Free Dynamic Downscaling Portal for Historic and Future Climate Data

Statistical downscaling of climate data has been widely described in the literature, with the aim of improving the reliability of local climatic parameters from coarse-resolution (often >20 km) global datasets. In this article, we present ClimateDT, a dynamic downscaling web tool for monthly historical and future time series at a global scale. The core of ClimateDT is the 1 km 1981–2010 climatology from CHELSA Climate (version 2.1), where the CRU-TS layers for the period 1901-current are overlayed to generate a historic time series. ClimateDT also provides future scenarios from CMIP5 using UKCP18 projections (rcp2.6 and rcp8.5) and CMIP6 using 5 GCMs, also available on the CHELSA website. The system can downscale the grids using a dynamic approach (scale-free) by computing a local environmental lapse rate for each location as an adjustment for spatial interpolation. Local predictions of temperature and precipitation obtained by ClimateDT were compared with climate time series assembled from 12,000 meteorological stations, and the Mean Absolute Error (MAE) and the explained variance (R2) were used as indicators of performance. The average MAEs for monthly values on the whole temporal scale (1901–2022) were around 1.26 °C for the maximum monthly temperature, 0.80 °C for the average monthly temperature, and 1.32 °C for the minimum monthly temperature. Regarding monthly total precipitation, the average MAE was 19 mm. As for the proportion of variance explained, average R2 values were always greater than 0.95 for temperatures and around 0.70 for precipitation due to the different degrees of temporal autocorrelation of precipitation data across time and space, which makes the estimation more complex. The elevation adjustment resulted in very accurate estimates in mountainous regions and areas with complex topography and substantially improved the local climatic parameter estimations in the downscaling process. Since its first release in November 2022, more than 1300 submissions have been processed. It takes less than 2 min to calculate 45 locations and around 8 min for the full dataset (512 records).

Nominal wage patterns, monopsony, and labour market power in early modern England

AbstractRecords of long‐eighteenth‐century English wage rates exhibit almost absolute nominal rigidity over many decades, alongside significant dispersion between the wages paid by different organizations for the same type of work in the same location. These features of preindustrial wages have been obscured by data aggregation and the construction of real wage series, which introduce variation. In this paper, we argue that the standard explanations for wage rigidity in economic history are insufficient. We show econometric evidence for monopsony power in one major organization and argue that the main historical wage series are also affected by employer power. Eighteenth‐century England had an imperfectly competitive labour market with large frictions. This gave large organizations the power to set wage policies. We discuss the implications for the eighteenth‐century British economy and research into long‐run wages more generally.

Predicting and Analyzing the Bitcoin's Price with the ARIMA Model

Virtual currencies have become one of the most welcomed financial investment properties by investors, managers, and researchers. In this essay, we analyze the Bitcoin price from 2017 to 2023 and forecast the price of the Bitcoin. ARIMA model plays an important role in time series prediction, and many researchers choose to use this model to analyze and forecast historical time series. Therefore, this paper is based on the Autoregressive Integrated Moving Average Model (ARIMA) mode, by using the previous Bitcoin price to forecast the future price of Bitcoin. The outbreak of COVID-19 in late 2019 which then spread from China to the world has had a huge negative impact on the global economy. At the same time, owing to the impact of COVID-19 being significant, this paper will also analyze the influence of the pandemic’s influence on Bitcoin. The result shows that the epidemic has a great influence on it and reflect a different consequence of other industry. Investors and others should be cautious about investing in bitcoin.

Research on Analysis and Optimization Strategies of Empresses in the Palace

This research delves into the intricacies of marketing strategies employed for the Chinese historical drama series “Empresses in the Palace." This study reveals vital insights through a comprehensive analysis of the existing marketing approaches and the unique challenges the historical drama genre poses. The “Empresses in the Palace” marketing strategies demonstrated innovation and creativity against intense competition within the historical drama landscape. Bridging the historical-contemporary gap and sustaining viewer engagement emerged as prominent challenges. This research has proposed various strategies tailored to address these challenges effectively. These strategies encompass diversified trailers, interactive engagement initiatives, and the integration of augmented reality experiences, aiming to optimize marketing endeavors. They seek to captivate contemporary audiences and maintain their interest throughout the series. The significance of the findings extends beyond this singular case, offering valuable lessons for historical drama marketing and its broader implications for the entertainment industry. This research recognizes the need for ongoing adaptation and innovation in engaging audiences with captivating historical narratives.

Evolution of transplant oncology indications: a single-institution experience over 40 years

Liver transplantation (LT) for uncommon tumoral indications has changed across the decades, with impaired results reported in the first historical series mainly for non-tumoral-related causes. Recently, renewed interest in liver transplant oncology has been reported. The study aims to analyze a mono-center experience exploring the evolution and the impact on patient survival of LT in uncommon tumoral indications. A retrospective analysis of 851 LT performed during 1982–2023 was investigated. 33/851 (3.9%) uncommon tumoral indications were reported: hepatocellular carcinoma (HCC) on non-cirrhotic liver (n = 14), peri-hilar (phCCA) (n = 8) and intrahepatic cholangiocarcinoma (i-CCA) (n = 3), metastatic disease (n = 4), hepatic hemangioendothelioma (n = 2), and benign tumor (n = 2). Uncommon tumoral indications were mainly transplanted during the period 1982–1989, with a complete disappearance after the year 2000 and a slight rise in the last years. Poor outcomes were reported: 5-year survival rates were 28.6%, 25.0%, 0%, and 0% in the case of HCC on non-cirrhotic liver, phCCA, i-CCA, and metastases, respectively. However, the cause of patient death was often related to non-tumoral conditions. LT for uncommon oncological diseases has increased worldwide in recent decades. Historical series report poor survival outcomes despite more recent data showing promising results. Hence, the decision to transplant these patients should be under the risk and overall benefit of the patient. The results of the ongoing protocol studies are expected to confirm the validity of the unconventional tumor indications.

AGE-GRADE DISTORTION

This article, which has the theme of age-grade distortion, presents the results of a documentary research and a qualitative approach, with the aim of analyzing the age-grade distortion rates of students in the final years of elementary school in the state network of Roraima and the actions implemented by the Department of Education and Sports (SEED) to meet this demand. The collection of data on school flow indicators was carried out in documents published by INEP and QEdu, with an analysis of the historical series of the School Census of the years 2017, 2018 and 2019 and information on the actions implemented by SEED, to assist students that make up this group, were obtained from publications on the official websites linked to that body. Results show that, although the age-grade distortion indicator shows a reduction, it is still present in both urban and rural state schools in the 15 municipalities of Roraima. Notwithstanding, the collection of data on the actions implemented by SEED has been limited to electronic media disclosures, it was found that actions are developed to reduce dropout, repetition and dropping out of school, such as the active search of students, the realization workshops and training courses for teachers working in flow correction classes.

HWWI Commodity Price Index: A Technical Documentation of the 2023 Revision

Abstract The HWWI Commodity Price Index, which aims at reflecting price developments of raw materials imported by highly industrialized countries, dates back to the 1950s. This paper explains the calculation of the index at the example of its latest revision, which became effective on 1st September 2023. While the index and its subindices are still calculated for OECD and European Momentary Union countries at four different frequencies (annual, monthly, quarterly and daily) for two different currencies (USD, EUR), it now additionally includes index values for several weighting periods (long-term: 2012–2021, pre-crisis: 2017–2019 and crisis years: 2020–2021), for individual countries and for different perspectives (daily revised, real-time and historical time series). The paper also provides detailed information on the employed methodology and the input data (commodities, import data, prices) used to calculate the weights and the index values. Although the revision is not without effect on the resulting index values, especially throughout the crises periods, the correlation between the initial and the revised index remains very high.

Application of the SMAP monthly model to the Piauitinga river basin located in the State of Sergipe, Brazil

The generation of flow data allows the assessment of the capacity to meet water demands, predict floods, and estimate the potential for hydraulic exploitation for electrical energy generation. In Brazil, precipitation data series, due to their ease of measurement, are more extensive compared to flow data series, enabling the use of hydrological models called rainfall-runoff models capable of estimating flows from precipitation data. Therefore, utilizing the Brazil Gridded Meteorological Data (BR-DWGD) database, this study aims to generate, calibrate, and validate flow data for the Piauitinga river basin located in the state of Sergipe, Brazil, using the monthly rainfall-runoff SMAP model. The soil parameters considered in the validation for the studied region showed a good fit to the observed data, achieving a Nash-Sutcliffe (NS) of 84% and log-Nash-Sutcliffe (NSLog) of 85% in calibration, and Nash-Sutcliffe (NS) of 70% and a log-Nash-Sutcliffe (NSLog) of 80% in validation. Therefore, since the rainfall-runoff model used exhibited good performance for the studied hydrographic basin, it becomes feasible to use the generated synthetic series to fill possible gaps in the historical series of monthly average flows.

Phenological Evaluation of Minority Grape Varieties in the Wine Region of Madrid as a Strategy for Adaptation to Climate Change

In this study, a total of 34 Spanish minority varieties were studied during four seasons from 2020 to 2023, and their behavior was characterized according to their main phenological stages (bud break, bloom, veraison, and maturity) and complete cycle. We focused on the varieties prospected in the central Spanish plateau and conserved in the “El Encín” grapevine collection, aiming to identify the potential for adaptation of these varieties and for them to be considered by winegrowers as an alternative to current climate change conditions. The growing degree days required for the expression of the phenological stage, and the duration of each stage, were compiled. Characteristics of oenological interest were also monitored, such as °Brix, pH, and titratable acidity in must at the time of harvest. This study was carried out in years with atypical snowfall and cold spells (winter 2021), as well as with heat waves (summer 2022), with average temperatures 3–5 °C higher than normal and absolute maximum temperatures over 40 °C. Both cases also exceeded records of historical series (1957–2019). Veraison has been identified as the stage most susceptible to damage from high temperatures, in addition to the maturation and duration of the complete cycle. The varieties were classified into five groups according to the duration of each phenological stage (very early, early, medium, late, and very late). Some varieties with late or very late maturation and with must characteristics of 20–23 °Brix, 3–5 g/L of titratable acidity, and pH 3.5–4.5, as well as others, retained stability in their phenological periodicity and must quality. The results suggest that special attention should be given to thoroughly evaluating these varieties, and that the strategies contemplated should be considered as a feasible cultivation alternative in viticulture to mitigate the effects of climate change.

Using local knowledge to reconstruct climate‐mediated changes in disease dynamics and yield—A case study on Arabica coffee in its native range

Societal Impact StatementAdapting agriculture to climate change requires an understanding of the long‐term relationship between climate, disease dynamics, and yield. While some countries have monitored major crop diseases for decades or centuries, comparable data is scarce or non‐existent for many countries that are most vulnerable to climate change. For this, a novel approach was developed to reconstruct climate‐mediated changes in disease dynamics and yield. Here, a case study on Arabica coffee in its area of origin demonstrates how to combine local knowledge, climate data, and spatial field surveys to reconstruct disease and yield time series and to postulate and test hypotheses for climate–disease–yield relationships.Summary While some countries have monitored crop diseases for several decades or centuries, other countries have very limited historical time series. In such areas, we lack data on long‐term patterns and drivers of disease dynamics, which is important for developing climate‐resilient disease management strategies. We adopted a novel approach, combining local knowledge, climate data, and spatial field surveys to understand long‐term climate‐mediated changes in disease dynamics in coffee agroforestry systems. For this, we worked with 58 smallholder farmers in southwestern Ethiopia, the area of origin of Arabica coffee. The majority of farmers perceived an increase in coffee leaf rust and a decrease in coffee berry disease, whereas perceptions of changes in coffee wilt disease and Armillaria root rot were highly variable among farmers. Climate data supported farmers' understanding of the climatic drivers (increased temperature, less rainy days) of these changes. Temporal disease‐climate relationships were matched by spatial disease‐climate relationships, as expected with space‐for‐time substitution. Understanding long‐term disease dynamics and yield is crucial to adapt disease management to climate change. Our study demonstrates how to combine local knowledge, climate data and spatial field surveys to reconstruct disease time series and postulate hypotheses for disease‐climate relationships in areas where few long‐term time series exist.

Hypersensitivity pneumonitis: application of a new diagnostic algorithm to a time series of the disease

ABSTRACT Background The diagnostic criteria for Hypersensitivity pneumonitis (HP) have changed over time. Our aim is to apply a recent diagnostic algorithm to a historical series of patients diagnosed with HP to assess its distribution according to current diagnostic criteria and the diagnostic confidence achieved. Research design and methods Application to each patient the algorithm criteria. The diagnosis was HP (≥90%), provisional high (70–89%) or low confidence (51–69%) or non-HP (unlikely) (≤50%); or HP, provisional or non-HP, if they had lung biopsy. Results 129 patients [mean age 64 ± 12 years; 79 (61.2%) women] were included of which 16 (12.4%) were diagnosed on the basis of high clinical suspicion. After applying the algorithm, 106 patients (82.2%) could be evaluated and 83 (78.3%) had a diagnosis of HP or high confidence. Lung biopsy was able to establish a diagnosis of certainty in another 21 patients and a provisional diagnosis in 9 more [total, 113 (87.6%)]. The 16 patients without strict diagnostic criteria for HP had a low confidence diagnosis. A total of 56 lung biopsies (64.4%) could have been avoided according to the new guidelines. Conclusions The application of this algorithm achieves a high diagnostic yield in HP, significantly reducing the number of lung biopsies required.

Outcomes of the transformation of follicular lymphoma to diffuse large B-cell lymphoma in the rituximab era: A population-based study.

Histological transformation (HT) to diffuse large B-cell lymphoma (DLBCL) is a common complication of follicular lymphoma (FL) and is usually associated with a dismal outcome. However, the survival rate of these patients has improved over the last 20 years with the introduction of rituximab. This study aimed to access the outcome of transformation to DLBCL (t-DLBCL) from FL in a retrospective series that began after the widespread use of rituximab use. In addition, we also compared survival between t-DLBCL and primary DLBCL (p-DLBCL) in the same timeframe. We utilized the Surveillance, Epidemiology, and End Results (SEER) database to identify patients with primary FL and patients with p-DLBCL between 2000 and 2020. Patients who had a subsequent diagnosis of DLBCL at least 2 months after FL diagnosis were identified as t-DLBCL. Finally, we identified 50,332 FL and 95,933 p-DLBCL. With a median follow-up of 119 months, 1631 patients developed t-DLBCL. The median time from FL diagnosis to t-DLBCL was approximately 4 years. The post-transformation survival (PTS) rate at 5 years was 49.6%, with a median PTS of 56 months. Older age, advanced stage, and early transformation were associated with worse PTS. Furthermore, t-DLBCL receiving chemotherapy or combined modality as initial therapy before HT was also associated with worse PTS, while the result was inverse when taking the impact of initial management strategy at HT into account. Taking t-DLBCL and p-DLBCL as a whole, comparable survival was observed between p-DLBCL and t-DLBCL receiving radiation or watch-and-wait as initial therapy prior to HT. The outcome of t-DLBCL in the rituximab era was better than historical series before the rituximab era. Due to the good prognosis, we did not recommend autologous stem cell transplantation for t-DLBCL receiving watch-and-wait or radiation as initial therapy before HT.

Analysis of hydraulic losses in the series from 1990 to 2022 at the Retiro small hydroelectric power plant

Hydraulic losses are a crucial variable in hydroelectric ventures as they can cause significant reductions in power generation. This article analyses the impact of hydraulic losses on the Retiro Small Hydroelectric Power Plant (SHPP), investigating their effect on monthly average power and, consequently, the efficiency of electricity generation. This study examines the historical series of load losses at the Retiro SHPP from 1990 to 2022. The calculations are based on flow data available in HidroWeb. The study considered the maximum and minimum flow rates in the historical flow series as constraints for hydropower generation. We used a multivariable function to calculate the efficiency of the hydraulic turbine, relating the turbine flow rate and the net water head. We developed mathematical relationships for head losses that occur in the grid, at the water intake, and due to friction from the intake to the turbine of the Retiro SHPP. The article presents a comparison between the actual monthly average power of the Small Hydroelectric Power Plant (SHPP) and the simulated monthly average power. We normalized the data for turbine flow and load loss to conduct statistical analysis. Kernel probability density was applied to understand the distribution shape of the data. Findings show that average monthly capacity is lowest in September, at approximately 0.81 MW. In March, the highest power occurs, approximately 14.19 MW. During the high flow period, the simulated average power, which accounts for load losses, closely matched the actual average power generated at the Retiro SHPP. In the months from July to October, despite being the period with the lowest head losses, it is the time when there is a greater opportunity to maximize energy generation at the power plant. An inefficient power generation system experiences significant load losses during specific periods of the year. To minimize this effect, it is crucial to understand the behavior of hydraulic losses and consider implementing mitigating measures.

Analyzing land use and land cover change in the Pra River Basin: A multi-tool approach for informed decision-making

Quantifying land use and land cover change (LULCC) is essential for addressing societal challenges and conserving ecosystems. This study focuses on analyzing the dynamics of five LULC classifications at a 30 m spatial resolution using satellite imagery from Landsat, coupled with various tools (Google Earth Engine, ArcMap, QGIS, and IDRISI SELVA) and indicators such as the Bare Soil Index (BSI). Historic time series data (2007, 2015, and 2023) were utilized to assess land change and predict future scenarios (2030 and 2063) using (CA) Markov models of the land change modeler (LCM). With a kappa coefficient and overall accuracy of 95 %, and 93 % respectively, our analysis reveals that LUC has significantly impacted nearly a third of the Pra River Basin (PRB) land area in two decades (2007–2023) after the launch of Agenda 2063 (The Africa We Want), with projected future impacts doubling under a business-as-usual (BAU) scenario. Geographical variations in LULC change processes are evident, with farmland abandonment, deforestation, and agricultural expansion observed predominantly in the northwestern region (Upper Pra, Anum, and Birim) of the basin. A 25 % increment of cultivated/farmland from 2007 to 2015 and a further 8 % in 2023. The northeast region (Upper and Lower Offin, Twifo Praso) faces deforestation and encroachment along riverbanks, along with built-up expansion due to mining activities. The northern region (Oda and Upper Pra) experiences exponential built-up and bare land increment. Accelerated phases of LULC change are attributed to factors such as mining (illegal and small-scale), agricultural production, migration, and population increment highlighting the influence of economic activities on land use dynamics. The concept gives a hybrid approach of using multi-tools in offering critical awareness of the state of the river system and encouraging the need for informed decision-making towards the anthropogenic impacts on the basin through effective management of LULC classes.

Unbanked and impoverished? Exploring banking and poverty interactions over time

AbstractIn 2019, the Federal Deposit Insurance Corporation (FDIC) reported that 7.1 million households in the United States (5.4%) were unbanked and lacked a checking or savings account). Using three leading household surveys, this paper documents how the interaction between bank access and poverty has evolved over time. We present a historical time series of unbanked rates, showing high‐unbanked rates for those in poverty even with increases in financial access over time. In the 1980s, 49.6% of households in poverty were unbanked while 22.8% were unbanked in 2019. Unbanked rates were even higher for Black and Hispanic households that were in poverty. In the 1980s, these groups had unbanked rates of 73.6% and 66.5% which declined to 38.4% and 31.8% in 2019, respectively. To explain differences in banking rates by race, we use binary Kitagawa‐Oaxaca‐Blinder decompositions. Socio‐demographic characteristics explain less than half the difference in unbanked rates for Blacks and around half for Hispanics.

Data-Driven Techniques for Short-Term Electricity Price Forecasting through Novel Deep Learning Approaches with Attention Mechanisms

The electricity market is constantly evolving, being driven by factors such as market liberalization, the increasing use of renewable energy sources (RESs), and various economic and political influences. These dynamics make it challenging to predict wholesale electricity prices. Accurate short-term forecasting is crucial to maintaining system balance and addressing anomalies such as negative prices and deviations from predictions. This paper investigates short-term electricity price forecasting using historical time series data and employs advanced deep learning algorithms. First, four deep learning models are implemented and proposed, which are a convolutional neural network (CNN) with an integrated attention mechanism, a hybrid CNN followed by a gated recurrent unit model (CNN-GRU) with an attention mechanism, and two ensemble learning models, which are a soft voting ensemble and a stacking ensemble model. Also, the optimized version of a transformer model, the Multi-Head Attention model, is introduced. Finally, the perceptron model is used as a benchmark for comparison. Our results show excellent prediction accuracy, particularly in the hybrid CNN-GRU model with attention, thereby achieving a mean absolute percentage error (MAPE) of 6.333%. The soft voting ensemble model and the Multi-Head Attention model also performed well, with MAPEs of 6.125% and 6.889%, respectively. These findings are significant, as previous studies have not shown high performance with transformer models and attention mechanisms. The presented results offer promising insights for future research in this field.