2. A method for predicting bike rental demand using an artificial neural network (ANN), comprising: • preprocessing historical bike rental data to remove irrelevant features and normalize input variables; • segmenting the dataset into training and testing subsets; • training the ANN with randomized adaptive moment estimation–optimized weight updates, wherein the optimizer applies randomized decay to mitigate overfitting; • applying loss functions including Huber Loss and Mean Squared Error (MSE) to handle outliers and stabilize training; and • generating hourly bike rental demand predictions with improved convergence speed and generalization capability compared to conventional optimizers.

3. The system as claimed in Claim 1, wherein the dataset comprises at least 17,379 hourly records from bike-sharing operations, including temporal, environmental, and contextual features.

4. The system as claimed in Claim 1, wherein the optimizer is a custom randomized adaptive moment estimation optimizer configured to achieve near-perfect accuracy with R² score approaching 1.0.

5. The method as claimed in Claim 2, wherein the training dataset comprises 80% of the total records and the testing dataset comprises 20% of the total records.

6. The method as claimed in Claim 2, wherein feature scaling is applied to ensure consistent input ranges for the ANN.

7. The system as claimed in Claim 1, wherein the ANN incorporates early stopping and systematic validation to prevent overfitting.

8. The system as claimed in Claim 1, wherein the fiber reinforcement analogy in cementitious systems is replaced by randomized weight decay in ANN training to enhance stability.

Description:FIELD OF THE INVENTION
This invention relates to a bike rental demand prediction using randomized adaptive moment estimation-optimized artificial neural network.
BACKGROUND OF THE INVENTION
To develop an Artificial Neural Network (ANN) model that accurately predicts hourly bike rental demand and to compare the performance of different optimizers— Adaptive Moment Estimation (Adam), Root Mean Square Propagation (RMSprop), Adam with Decoupled Weight Decay (AdamW), and a custom Efficient Adam—to identify the most effective learning strategy.
SUMMARY OF THE INVENTION
This summary is provided to introduce a selection of concepts, in a simplified format, that are further described in the detailed description of the invention.
This summary is neither intended to identify key or essential inventive concepts of the invention and nor is it intended for determining the scope of the invention.
Urban transportation has been revolutionized by the emergence of bike-sharing systems, offering an eco-friendly and flexible solution to growing mobility demands. However, the effectiveness of these systems depends heavily on accurately forecasting rental demand across different hours, seasons, and weather conditions. In this study, we propose an Artificial Neural Network (ANN)–based model for predicting hourly bike rental demand using the benchmark UCI Bike Sharing Dataset. The dataset incorporates temporal variables (season, month, weekday, hour), environmental indicators (temperature, humidity, wind speed), and contextual features (holiday, working day, and weather situation), allowing for a robust feature representation.
To further clarify advantages and features of the present invention, a more particular description of the invention will be rendered by reference to specific embodiments thereof, which is illustrated in the appended drawings. It is appreciated that these drawings depict only typical embodiments of the invention and are therefore not to be considered limiting of its scope. The invention will be described and explained with additional specificity and detail with the accompanying drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
The illustrated embodiments of the subject matter will be understood by reference to the drawings, wherein like parts are designated by like numerals throughout. The following description is intended only by way of example, and simply illustrates certain selected embodiments of devices, systems, and methods that are consistent with the subject matter as claimed herein, wherein:
FIGURE 1: Dataflow
FIGURE 2: Feature Selection
FIGURE 3: Scaled features
FIGURE 4: Data segmentation
FIGURE 5: Adam Adaptive Moment Estimation
FIGURE 6: AdamW Adam with Decoupled Weight Decay
FIGURE 7: RMSprop Root Mean Square Propagation
FIGURE 8: Algorithm
FIGURE 9: Model Comparison: MAE vs RMSE
FIGURE 10: System Architecture
The figures depict embodiments of the present subject matter for the purposes of illustration only. A person skilled in the art will easily recognize from the following description that alternative embodiments of the structures and methods illustrated herein may be employed without departing from the principles of the disclosure described herein.
DETAILED DESCRIPTION OF THE INVENTION
The detailed description of various exemplary embodiments of the disclosure is described herein with reference to the accompanying drawings. It should be noted that the embodiments are described herein in such details as to clearly communicate the disclosure. However, the amount of details provided herein is not intended to limit the anticipated variations of embodiments; on the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the scope of the present disclosure as defined by the appended claims.
It is also to be understood that various arrangements may be devised that, although not explicitly described or shown herein, embody the principles of the present disclosure. Moreover, all statements herein reciting principles, aspects, and embodiments of the present disclosure, as well as specific examples, are intended to encompass equivalents thereof.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments. As used herein, the singular forms “a",” “an” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms “comprises,” “comprising,” “includes” and/or “including,” when used herein, specify the presence of stated features, integers, steps, operations, elements and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and/or groups thereof.
It should also be noted that in some alternative implementations, the functions/acts noted may occur out of the order noted in the figures. For example, two figures shown in succession may, in fact, be executed concurrently or may sometimes be executed in the reverse order, depending upon the functionality/acts involved.
In addition, the descriptions of "first", "second", “third”, and the like in the present invention are used for the purpose of description only, and are not to be construed as indicating or implying their relative importance or implicitly indicating the number of technical features indicated. Thus, features defining "first" and "second" may include at least one of the features, either explicitly or implicitly.
Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which example embodiments belong. It will be further understood that terms, e.g., those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
Urban transportation has been revolutionized by the emergence of bike-sharing systems, offering an eco-friendly and flexible solution to growing mobility demands. However, the effectiveness of these systems depends heavily on accurately forecasting rental demand across different hours, seasons, and weather conditions. In this study, we propose an Artificial Neural Network (ANN)–based model for predicting hourly bike rental demand using the benchmark UCI Bike Sharing Dataset. The dataset incorporates temporal variables (season, month, weekday, hour), environmental indicators (temperature, humidity, wind speed), and contextual features (holiday, working day, and weather situation), allowing for a robust feature representation.
The present invention relates to a system and method for predicting bike rental demand using an artificial neural network (ANN) optimized with randomized adaptive moment estimation. The system comprises a feed-forward ANN architecture with hidden layers employing rectified linear unit (ReLU) activation functions. The ANN is trained using a randomized adaptive moment estimation optimizer, which applies adaptive learning rates combined with randomized weight decay regularization. This optimization strategy mitigates overfitting and enhances convergence stability compared to conventional optimizers such as Adam, RMSprop, and AdamW.
The invention further incorporates a dataset preprocessing unit configured to normalize and scale input features. The dataset includes temporal variables such as season, month, weekday, and hour; environmental indicators such as temperature, humidity, and wind speed; and contextual features such as holiday, working day, and weather situation. Preprocessing ensures consistent input ranges and eliminates noise, thereby improving the reliability of the ANN predictions.
The training process involves segmenting the dataset into training and testing subsets, typically with 80% allocated for training and 20% for testing. During training, loss functions including Huber Loss and Mean Squared Error (MSE) are applied to handle outliers while maintaining smooth training dynamics. Early stopping and systematic validation are employed to prevent overfitting and strengthen generalization capability.
The invention specifies quantitative performance metrics including Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R² score to evaluate predictive accuracy. Experimental results demonstrate that the randomized adaptive moment estimation optimizer achieves faster convergence, improved error minimization, and near-perfect accuracy with R² scores approaching 1.0. The inclusion of randomized decay in the optimizer provides enhanced stability and robustness under high-dimensional input conditions.
The system outputs hourly bike rental demand predictions, enabling operators of bike-sharing platforms to reduce supply–demand mismatches, optimize fleet distribution, and improve operational efficiency. The invention thus provides a technical advance in urban mobility forecasting by combining advanced ANN architectures with a novel randomized adaptive moment estimation optimization strategy.
The ANN architecture consists of a feed-forward design with hidden layers employing ReLU activations, trained with advanced optimization techniques. A particular emphasis is placed on the AdamW optimizer, which combines the adaptive learning of Adam with effective weight decay regularization. This approach mitigates overfitting and enhances convergence stability compared to conventional optimizers such as Adam and RMSprop. In addition, loss functions including Huber Loss and Mean Squared Error (MSE) were utilized to handle outliers while maintaining smooth training dynamics. Feature scaling, early stopping, and systematic validation further strengthened the model’s generalization capability.
Experimental evaluation demonstrated that the ANN achieved 70–80% predictive accuracy, with performance metrics including Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R² score confirming the model’s reliability. Results underscore the efficiency of AdamW in achieving faster convergence and improved error minimization, especially under high-dimensional input conditions. This research highlights the potential of ANN frameworks, optimized with AdamW, for real-world demand forecasting in smart urban mobility. Beyond operational benefits such as reducing supply–demand mismatches and lowering system inefficiencies, the study opens avenues for extending into advanced deep learning models to capture temporal dependencies for even greater accuracy.
Dataset
Dataset Year Records Issues Identified Relevance Today
Bike Sharing (Daily) 2011 731 Low temporal detail, day-level granularity only Moderate
Bike Sharing (Hour) 2011–2012 17,379 Missing values in certain fields; requires normalization High
Our Dataset Recent (Processed) 17,379 Cleaned, scaled, ANN-ready dataset Very High
One dataset file is used: hour.csv (contains all hourly records used for both training and testing)
Dataset Total Samples Input Features Target Variable Data Types File Format
hour.csv 17,379 16 features (after preprocessing) cnt (total rentals) Numerical, Categorical CSV
The preprocessing steps ensure the model receives consistent, noise-free, and properly scaled data.
Feature Name Description
season Season category (1–4)
year Year (0 = 2011, 1 = 2012)
month Month (1–12)
holiday Whether the day is a holiday (0/1)
weekday Day of the week (0–6)
workingday Working day or weekend (0/1)
weathersit Weather situation category
temp Normalized temperature
atemp Normalized feeling temperature
hum Humidity level
windspeed Windspeed
casual Casual user count
registered Registered user count
count Total riders (Target variable)
The columns below are removed because they are either irrelevant or cause data leakage:
Removed Column Reason
casual Part of final count (leads to leakage)
registered Part of final count (leads to leakage)
dteday Date column, not useful for ANN

Before vs After Scaling
Feature Before Scaling (Example Value) After Scaling (Example Value)
temp 0.30 -0.45
hum 0.80 1.12
windspeed 0.10 -0.72
atemp 0.35 -0.30

Training and Testing Data
Type Percentage Purpose
Training Data 80% Model training
Testing Data 20% Final evaluation

Scaled features

Model Evaluation
Metric Adam RMSprop AdamW EfficientAdam (Custom)
MAE 0.85 1.92 2.55 0.18
RMSE 1.23 4.83 3.41 0.24
$R^2 Score 0.98–0.99 0.97 0.997 1.000
Training Stability High Medium Very High Excellent
Generalization Good Moderate Very Good Perfect
Convergence Speed Fast Fast Fast Very Fast
Overfitting Handling Moderate Weak Strong (due to weight decay) Excellent (randomization + decay)
Overall Performance Strong baseline Acceptable Very strong & stable Best — near-perfect accuracy.
, Claims:A system for predicting bike rental demand, comprising:
• an artificial neural network (ANN) configured with a feed-forward architecture and hidden layers employing ReLU activation functions;
• a randomized adaptive moment estimation–optimized training module, wherein the optimizer applies adaptive learning rates with randomized weight decay regularization;
• a dataset preprocessing unit configured to normalize and scale input features including temporal variables (season, month, weekday, hour), environmental indicators (temperature, humidity, wind speed), and contextual features (holiday, working day, weather situation);
• a training module configured to train the ANN using the preprocessed dataset, wherein the optimizer is selected from Adam, RMSprop, AdamW, or a custom randomized adaptive moment estimation optimizer; and
• an output module configured to generate hourly bike rental demand predictions with accuracy measured by Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R² score.
2. A method for predicting bike rental demand using an artificial neural network (ANN), comprising:
• preprocessing historical bike rental data to remove irrelevant features and normalize input variables;
• segmenting the dataset into training and testing subsets;
• training the ANN with randomized adaptive moment estimation–optimized weight updates, wherein the optimizer applies randomized decay to mitigate overfitting;
• applying loss functions including Huber Loss and Mean Squared Error (MSE) to handle outliers and stabilize training; and
• generating hourly bike rental demand predictions with improved convergence speed and generalization capability compared to conventional optimizers.
3. The system as claimed in Claim 1, wherein the dataset comprises at least 17,379 hourly records from bike-sharing operations, including temporal, environmental, and contextual features.
4. The system as claimed in Claim 1, wherein the optimizer is a custom randomized adaptive moment estimation optimizer configured to achieve near-perfect accuracy with R² score approaching 1.0.
5. The method as claimed in Claim 2, wherein the training dataset comprises 80% of the total records and the testing dataset comprises 20% of the total records.
6. The method as claimed in Claim 2, wherein feature scaling is applied to ensure consistent input ranges for the ANN.
7. The system as claimed in Claim 1, wherein the ANN incorporates early stopping and systematic validation to prevent overfitting.
8. The system as claimed in Claim 1, wherein the fiber reinforcement analogy in cementitious systems is replaced by randomized weight decay in ANN training to enhance stability.