1. bookAHEAD OF PRINT
Dettagli della rivista
License
Formato
Rivista
eISSN
2444-8656
Prima pubblicazione
01 Jan 2016
Frequenza di pubblicazione
2 volte all'anno
Lingue
Inglese
Accesso libero

Research on multi-dimensional optimisation design of user interface under Rhino/GH platform

Pubblicato online: 21 Oct 2022
Volume & Edizione: AHEAD OF PRINT
Pagine: -
Ricevuto: 14 May 2022
Accettato: 15 Jun 2022
Dettagli della rivista
License
Formato
Rivista
eISSN
2444-8656
Prima pubblicazione
01 Jan 2016
Frequenza di pubblicazione
2 volte all'anno
Lingue
Inglese
Introduction

User interface (UI) [1] is a medium for interaction and message exchange between people and electronic computer systems, and is a comprehensive environment for users to use electronic computers. At present, the definition of UI is relatively broad, and it is not restricted to the graphical UI of human–machine interaction. In a broad sense, UI is a collection of interaction methods between users and systems [2, 3]. These systems do not only refer to computer programs, but also include a specific machine, equipment, complex tool, etc. [4] The UI can be seen as representing a face-to-face information exchange method between people and computers, and the formation of the UI comes from the properties of artificial objects. The UI is a collection of interaction methods between users and the system, and is also the software and hardware part of the electronic computer system that realises the exchange of information between the user and the computer [5]. As shown in Figure 1, the UI is divided into a hardware interface and a software interface. The hardware interface mainly refers to the computer keyboard and mouse, as shown in Figure 1(a), and the software interface mainly refers to the interface for direct information exchange between the user and the computer, as shown in Figure 1(b) and (c), that is, the user browses multimedia through the computer network, and the text web interface. The purpose of the UI is to enable the user to operate the electronic computer system conveniently and efficiently to achieve two-way interaction. Nowadays, with the rapid development of computer networks, people have steadily increasing requirements for computers to process data, as well as handle other applications, including entertainment. The amount of software installed in the computer is increasing day by day, and this puts forward requirements on how to realise the interaction of the computer with a fast and comfortable interface.

Fig. 1

Simple schematic diagram of the interface

Currently, when evaluating the parameters required for the UI, we mainly consider inclusivity and usability. Since the main service object of the interactive interface is the user, the inclusive interactive interface designed by Microsoft research has been formulated in such a way that users can solve problems according to their own needs; the UI has been endowed with a capability to adapt to the diversity of individuals, and this approach has resulted in an easy-to-use, flexible and good user experience. Brandon Antonio Cárdenas-Sainz et al. [6] enhanced the user's interactive learning environment through a 3D virtual environment. Their research showed that through different types of multimedia (visual, hearing) and interactive ability (kinesthesis), using multiple regression analysis can improve the compatibility of the interactive interface. Moore [7] and others studied the impact of physical and social aspects on UI. They compared the manifestations of three different forms of VUI (non/physical/social), and the results showed that the external voice environment will affect the user's response to the interface. Miia Jansson et al. studied interface design, inefficiency, failure and negative impact of information retrieval on users. Their research showed that human–computer interaction requires a certain degree of efficiency, inclusiveness [8, 9] and reliability [10]. With regard to the field of human–computer interaction, Abascal and Nicolle [11] analysed the relationship between human–computer interaction and inclusivity. Their research pointed out that inclusivity mainly includes common problems in the design process, and suggested guidelines for improving accessibility. The inclusive design mainly includes five parts: problem identification, system prototype visualisation, matching system and user expectations, developing control and input methods, and evaluating system usability and accessibility. Langdon et al. [12] proposed a cognitive analysis framework and components, and established a cognitive interaction model to assist in the evaluation and analysis of inclusive design. As shown in Figure 2, the inclusiveness of the interactive interface is mainly reflected in the interactive interface between the user and the computer, and the quality of the interactive interface is mainly related to the user type and user ability level.

Fig. 2

Interactive interface inclusion indicators

For the usability of the interactive interface, the main point is that the computer system can stably and quickly switch to the interface required by the user in a specific situation or software interaction required by the user. Some scholars have proposed that the interactive interface should have several requirements, such as learnability, high efficiency, memorability, low error rate, and automatic repair of errors and satisfaction [13], and for users, it should have relatively simple interaction. The switching operation and system are common [14]. The evaluation of usability is mainly based on the efficiency, effectiveness and user satisfaction of interface interaction (as shown in Figure 3). The longer the response time to complete a task, the higher the efficiency; at the same time, the greater the number of interactive interfaces that are completed within the specified time frame, the faster the speed. For effectiveness, according to the error rate and degree of completion in the interactive interface, the lower the error rate, the higher the completion rate, and the better the effectiveness. In this paper, we choose the response time of a long-term and frequent switching of an interactive interface and the response time of repeated interaction of multiple interfaces as the main criteria for usability. Alomari [15] measures the utility of a UI by testing its usability and using UI and user-experience assessments; these are studies that can be used to evaluate the usefulness of e-learning environments, and they find widespread application in computer science and software engineering courses.

Fig. 3

Interactive interface usability indicators

However, when implementing interface generation, the commonly used software applications mainly include PS, AI.AE.C4D.XD.ARP and so on. PS is mainly used for image processing, AI icon design, AE dynamic effect design, and C4D is mainly used for modelling. For a single interface, there is a need for proper processing to enable the transition to the displayed interface. Nowadays, the interface is becoming increasingly complex. Yang and Horie [16] improved the UI by including neural network, which improved the accuracy of complex gestures. Yang et al. [17] optimised the UI based on recurrent neural network (RNN) and gesture sensor. The Rhino + Grasshopper platform gradually became popular in the design stage because of its agility and speed, and its great potential to achieve deep cooperation through development [18]. At the beginning, the designed interface can be analysed in various aspects such as structure and fluid [19, 20], and the performance of the scheme can be comprehensively evaluated. Therefore, this paper adopts the Rhino/GH platform to study the multi-dimensional optimisation of user exchange interface. This paper mainly studies the UI optimisation performance under the introduction of different algorithms, and determines the reliability and inclusiveness of the interactive interface by comparing different response times.

Theoretical analysis
Introduction to the theoretical work of Rhino/GH platform

The biggest working feature of Rhino is that it uses the non-spline curve modelling method [21] (i.e. NURBS modelling method). It has the advantages of high modulus accuracy and flexible use [22]. Compared with other types of modelling software, it has lower requirements on the application environment, and takes up less computer space due to its fast running speed. This advantage ensures that more interactive use of computer interfaces can be achieved under the same storage space. In addition, it can cooperate with other extension plug-ins to realise extension function and realise powerful extension by substituting parameterised design. To simplify some complex parts, parametric modelling is increasingly widely used in some complex designs and applications, and complex modelling analysis is completed with the help of parametric modelling. As Rhino's parametric design platform, Grasshopper can complete advanced controls such as visual image generation and transformation, and achieve certain effects on the surface of performance analysis. This provides not only a new imagination space for structural design but also new solutions for some complex designs. Compared with the traditional design modelling, Grasshopper has the following characteristics, as shown in Figure 4: compellability, visualisation, and extensibility. Compellability means that by writing a series of programs, it can be connected to external systems, and real-time parameter changes can be realised without causing mutation of the interactive interface, thereby ensuring the multi-dimensional stability of the interactive interface; visualisation can generate geometric models according to specified algorithms. Zhang et al. [23] formulate the algorithm using Rhino/Grasshopper and Python. The total load of the optimal scheme is 15.8% lower than that of the worst scheme, and 4.2% lower than that of the original scheme. Herrema et al. [24] demonstrated the IGA-based parametric design optimisation framework implemented by the Rhino/GH platform's access algorithm modelling interface, which improved the simplification and visualisation of their research content. Esfahani [25] optimised the solar gain of southern hemisphere roofs by introducing an algorithm. Therefore, in this paper, we introduce long short-term memory (LSTM) and gated recurrent unit (GRU) algorithms to generate relevant feedback information, and study their speed and reliability in improving the interactive interface switching system when processing data; the scalable performance realises the multi-parameter function of the interactive interface and improves the interactive interface quantity.

Fig. 4

GH characteristics

Parametric modelling of Rhino/GH platform

There are many GH auxiliary external plug-ins, which are mainly divided into the following categories, as shown in Figure 5: design, algorithm, structure, conversion, and interactive simulation; these are mentioned here as several common external plug-ins. By simulating some related physical phenomena such as gravity, the modulation of physical formulas in the system can be improved; the algorithm class is mainly for parameterisation and improving data processing speed. The structural class can be targeted to give force analysis to each specific location, etc.; the conversion class is mainly to realise multi-software intercommunication, realise the Internet of things, human–computer exchange, visualisation, etc.; interactive simulation is mainly used to cluster data simulation and realise data visual interaction.

Fig. 5

Types of GH external plug-ins

In the research work of this paper, LSTM and GRU algorithms are mainly used to optimise a model of the traditional interactive interface. The LSTM long-term and short-term neural network is evolved from the RNN [26]. The RNN adds weights to the neural network of the same layer, and introduces real-time feedback to correct the neural network. The LSTM neural network changes and processes the structure of the RNN based on RNN to solve the problem of its gradient disappearance [27]. This is mainly because the LSTM operates almost linearly, with the entire operation as linked rows. The long-term and short-term neural network is mainly composed of three control units: forget gate, input gate and output gate [28]. Its basic unit structure is shown in Figure 6. The main working formula is as follows: ft=σ(wf[ht1,xt]+bf) {f_t} = \sigma \left( {{w^f} \cdot \left[ {{h_{t - 1}},{x_t}} \right] + {b^f}}\right)

Fig. 6

Basic structure of LSTM. LSTM, long short-term memory

The formula for calculating the input gate is it=σ(wi[ht1,xt]+bi) {i_t} = \sigma \left( {{w^i} \cdot \left[ {{h_{t - 1}},{x_t}} \right] + {b^i}}\right) C˜t=tanh(wc[ht1,xt]+bc) {\tilde C_t} = \tanh \left( {{w^c} \cdot \left[ {{h_{t - 1}},{x_t}} \right] + {b^c}} \right) where bf represents the bias term, the activation function σ selects the sigmoid activation function, and Wf is the weight of the forget gate heavy matrix. The final output calculation formula is ot=σ(wo[ht1,xt]+bo) {o_t} = \sigma \left( {{w^o} \cdot \left[ {{h_{t - 1}},{x_t}} \right] + {b^o}}\right) ht=ot*tanh(Ct) {h_t} = {o_t} * \tanh \left( {{C_t}} \right)

GRUs is a gating mechanism in RNN [29]. Similar to LTMS network, GRU neural network is also a variation model of RNN network, but it is more concise than LTMS network. There are only update gates and reset gates in the network model [30, 31, 32]. The special significance of these two gating mechanisms is that the information in the sequence is preserved for a long time, and the information in the sequence is not cleared over time or removed because it is irrelevant to prediction [33]. The update gate formula is as follows: rt=σ(WrXt+Urht1+bz) r_{t} = \sigma \left( W_{r}X_{t} + U_{r}h_{t-1} + b_{z}\right) where Xt represents the input vector value at time t, the input sequence X represents the t component, the weight matrix is represented by Wz and Uz, respectively, bz represents the bias term, and ht−1 represents the information at time t − 1 in the previous step.

The main function of the reset gate is to determine the abandoned information, using the formulas: rt=σ(WrXt+Urht1+br) r_{t} = \sigma \left( W_{r}X_{t} + U_{r}h_{t-1} + b_{r}\right) ht.=tanh(WXt+rtUhr1) h_{t}^{.} = \text{tanh} \left( WX_{t} + r_{t} Uh_{r-1}\right) ht=(1Zt)ht1+Zth˜t h_{t} = \left( 1- Z_{t}\right) \cdot h_{t-1} + Z_{t} \cdot \tilde{h}_{t}

Z represents the activation result of the update gate, and the product of 1 − Zt and ht−1 represents the part of the final memory information retained at the previous moment.

Evaluation criteria

The stability and rapid performance of the UI are affected by many factors. The selection of evaluation indicators is important for the interactive interface, and the training of the model is very important. In the experiment, the mean square error function is selected as the target cost function, and the experimental model is trained.

The process aims to make the objective cost function increasingly smaller and the prediction accuracy increasingly higher [34, 35]. The mean squared error is taken as the target cost function, which can stabilise the model effectively. The mean square error calculation is shown in Eq. (10): RMSE=1ni=1n(yiyp)2 RMSE = \sqrt{\frac{1}{n} \sum\limits_{i=1}^{n} \left( y_{i} - y_{p}\right)^{2}}

For atrained model, the goodness of fit is used to determine the accuracy of the model. The closer the goodness of fit is to 1, the better the fit, and vice versa. The goodness of fit is calculated as in Eq. (11) R2(yi,yp)=1i=1n(yiyp)2i=1n(yiy¯)2 R^{2} \left( y_{i}, y_{p}\right) = 1 - \frac{\sum\limits_{i=1}^{n} \left( y_{i} - y_{p}\right)^{2}}{\sum\limits_{i=1}^{n} \left( y_{i} - \overline{y} \right)^{2}} where n is the number of samples, yi is the real value and yp is the predicted value. The smaller the gap between the predicted value and the real value, the better the fit.

The parameter selection of the model directly determines the effect of the model, through a large number of parameter adjustments. The specific parameters of the model selected in this paper are shown in Table 1.

Model parameter table

Parameter Layer Number of neurons Dropout Batch_size Learn rate Nb_epoch
Value 2 256 0.83 500 0.005 10

The target cost function uses the mean square error function to make the prediction more accurate. In order to avoid overfitting of the model, a dropout layer is added during the model training process, and the number of neurons in each layer is 256. In each round, 500 datasets are input for training, for a total of 10 rounds of training.

Analysis of experimental results

The experiment carried out multi-dimensional optimisation design for the reliability and inclusiveness of the interactive interface, and used LSTM and GRU algorithms to perform expansion, whereas Rhino/GH platform was used to enhance its data processing capability, ensure the reliability of the interactive interface, and avoid the occurrence of jamming and other phenomena. The experiment was carried out on Window10, 64-bit operating system, and the computer running memory was 16GB. The experiment selects a specific interface and performs 100 cycles of switching experiments on the same operating system to compare and analyse the impact time of the switching interfaces; in addition, to verify the inclusiveness of its interactive interfaces, we selected 20 interfaces on the same computer operating system, performed mutual switching of these 20 interfaces, and compared and analysed the response obtained when the interface is switched; additionally, to verify the reliability of the interactive interface, we used the same computer operating system. The number of interfaces was also compared.

Before conducting the experiment, we first conducted a comparative analysis of the LSTM and GRU models, analysed and compared the accuracy of the models and their root mean squares (RMSs), and selected the optimal algorithm suitable for the Rhino/GH platform expansion combination. The accuracy and loss of the model are shown in Table 2. It can be seen from Table 2 that the parameter accuracy of the LSTM and GRU algorithms can reach >98%. Using the GRU algorithm to process data on the Rhino/GH platform yields a higher relative accuracy, which can reach 99.3%, and the model training loss function is only 0.00212, which is 50% of that observed in the case of the LSTM algorithm. The main reason is that the GRU algorithm can update the data in real-time and train continuously during multi-interface interaction, thereby ensuring the usability and mutual inclusiveness of the interactive interface.

Model accuracy table

Model Loss acc

LSTM 0.00452 98.2%
GRU 0.00212 99.3%

GRU, Gated recurrent unit;

LSTM, long short-term memory

In order to evaluate the two algorithms from various aspects, we use RMS and goodness-of-fit analysis for further analysis and comparison, and the specific results are shown in Table 3.

RMS and fit table

Module RMSE R2

LSTM 0.983 0.835
GRU 0.879 0.910

GRU, Gated recurrent unit;

LSTM, long short-term memory;

RMSE, root mean square error

From the table, we can see that the RMS of GRU algorithm is smaller than that of LSTM, and the fitting degree is 91%, which is better than the 83.5% observed in the case of LSTM. Therefore, when optimising the interactive interface, choosing the GRU algorithm will be more realistic.

Figure 7(a) shows a comparison chart of the time required to switch between the traditional interactive interface and the Rhino/GH platform. In the actual test of 100 switching times, we can find that for the traditional interactive interface, the time is between 245 ms and 361 ms, and the fluctuation range is 47.3%. In contrast, the interactive interface based on the Rhino/GH platform yields excellent performance and high stability in the test. The interaction time is between 118 ms and 146 ms, and the maximum fluctuation range is only 23.7%; and compared with the traditional method, the decline rate is about 50%. In addition, the switching time has also dropped by >50% as a whole. In order to further verify the stability of the interactive interface of the Rhino/GH platform, we selected 20 interfaces to be online at the same time, and compared the change of the interaction response time with the results observed in the case of the traditional interaction method. The results are shown in 7(b). As can be seen from Figure 7(b), when multiple interface switching is frequent, the traditional interactive interface technology shows a gradual increasing trend with the increase of the number of interfaces, the response time is between 295 ms and 450 ms, and the fluctuation range is as high as 52.5%; and it shows an increasing trend with the increase of the interface. In addition, the overall response time is between 45 ms and 89 ms longer than the single interface's 245–361 ms. In contrast, the optimised interactive interface of Rhino/GH has a response time of 123–159 ms when 20 interfaces are switched, the maximum fluctuation range is 29.2%, and the response time remains within the stable range; additionally, the overall response time is only 13 ms longer than the 118–146 ms range of the monomer. The comprehensive results show that the optimised interactive interface of Rhino/GH has obvious usability, and the inclusiveness between different interfaces is good.

Fig. 7

Time response diagram of interactive interface (a) single interface interaction time response diagram; (b) multi-interface interaction time response diagram

As can be seen from Figure 8, with the increase of the number of interactive interfaces, the proportion of computer running storage gradually increases until it can no longer work. Based on the traditional interactive interface, when the number of interfaces reaches 27, the proportion of memory reaches 89.1%, at which point the computer begins to appear sluggish, inefficient and unstable. At this time, the proportion of memory based on Rhino/GH optimisation platform was 73.2%, a decrease of 15.9%. After optimising the interactive interface, when the number of interfaces reached 33, the proportion of memory reached 89.2%, and then the computer operating system began to increasingly present with the characteristics of sluggishness, inefficiency, etc. The optimised interactive interface can increase the usage rate of computer memory by 22.2%.

Fig. 8

The proportion of memory space

Conclusion

The interactive interface optimisation research of Rhino/GH platform optimisation proposed in this paper introduces LSTM and GRU algorithms to optimise the platform data processing, and compares the long-term frequent switching of single interface and multi-interface interaction with the traditional interactive interface. The main conclusions are as follows:

When the GRU algorithm is added to the Rhino/GH platform for data processing, the model accuracy can reach 99.3%, and the RMS remains at 0.879. The introduction of GRU can realise real-time update of data, and avoid slow response caused by untimely processing of interactive interface data when the amount of data is too large, thereby affecting the use effect.

We introduce the interactive interface optimised by the Rhino/GH platform. When a single interface is frequently interacted for a long time, the overall response time is 50%, which is a similarity observed with the traditional interactive interface; again, the time fluctuation is within 23.7%, which is approximately 23.6% lower than the 47.3% observed in the case of the traditional interactive interface.

In the case of multi-interface interaction, the traditional interactive interface shows a trend of increasing response time as the interactive interface increases, and the fluctuation range also increases to 52.5%, which means that the instability of the interactive interface increases. The interactive interface optimised by the Rhino/GH platform maintains a fluctuation range within 29.2%, and the time increases by 13 ms, with good stability.

The optimised interactive interface of Rhino/GH platform can increase the utilisation rate of storage space by 22.2% when it is in use.

Fig. 1

Simple schematic diagram of the interface
Simple schematic diagram of the interface

Fig. 2

Interactive interface inclusion indicators
Interactive interface inclusion indicators

Fig. 3

Interactive interface usability indicators
Interactive interface usability indicators

Fig. 4

GH characteristics
GH characteristics

Fig. 5

Types of GH external plug-ins
Types of GH external plug-ins

Fig. 6

Basic structure of LSTM. LSTM, long short-term memory
Basic structure of LSTM. LSTM, long short-term memory

Fig. 7

Time response diagram of interactive interface (a) single interface interaction time response diagram; (b) multi-interface interaction time response diagram
Time response diagram of interactive interface (a) single interface interaction time response diagram; (b) multi-interface interaction time response diagram

Fig. 8

The proportion of memory space
The proportion of memory space

Model parameter table

Parameter Layer Number of neurons Dropout Batch_size Learn rate Nb_epoch
Value 2 256 0.83 500 0.005 10

Model accuracy table

Model Loss acc

LSTM 0.00452 98.2%
GRU 0.00212 99.3%

RMS and fit table

Module RMSE R2

LSTM 0.983 0.835
GRU 0.879 0.910

Wachs, J.P., et al., Vision-based hand-gesture applications. Communications of the ACM, 2011. 54(2): p. 60–71. WachsJ.P. Vision-based hand-gesture applications Communications of the ACM 2011 54 2 60 71 10.1145/1897816.1897838 Search in Google Scholar

Ratcliffe, L. and S. Puthusserypady, Importance of Graphical User Interface in the design of P300 based Brain–Computer Interface systems. Computers in Biology and Medicine, 2020. 117: p. 103599. RatcliffeL. PuthusserypadyS. Importance of Graphical User Interface in the design of P300 based Brain–Computer Interface systems Computers in Biology and Medicine 2020 117 103599 10.1016/j.compbiomed.2019.10359932072963 Search in Google Scholar

Gaouar, L., et al., HCIDL: Human-computer interface description language for multi-target, multimodal, plastic user interfaces. Future Computing and Informatics Journal, 2018. 3(1): p. 110–130. GaouarL. HCIDL: Human-computer interface description language for multi-target, multimodal, plastic user interfaces Future Computing and Informatics Journal 2018 3 1 110 130 10.1016/j.fcij.2018.02.001 Search in Google Scholar

Mulfari, D., A. Celesti and M. Villari, A computer system architecture providing a user-friendly man machine interface for accessing assistive technology in cloud computing. Journal of Systems and Software, 2015. 100: p. 129–138. MulfariD. CelestiA. VillariM. A computer system architecture providing a user-friendly man machine interface for accessing assistive technology in cloud computing Journal of Systems and Software 2015 100 129 138 10.1016/j.jss.2014.10.035 Search in Google Scholar

Messaoud, S., et al., Deep convolutional neural networks-based Hardware–Software on-chip system for computer vision application. Computers & Electrical Engineering, 2022. 98: p. 107671. MessaoudS. Deep convolutional neural networks-based Hardware–Software on-chip system for computer vision application Computers & Electrical Engineering 2022 98 107671 10.1016/j.compeleceng.2021.107671 Search in Google Scholar

Cárdenas-Sainz, B.A., et al., Integration and acceptance of Natural User Interfaces for interactive learning environments. International Journal of Child-Computer Interaction, 2022. 31: p. 100381. Cárdenas-SainzB.A. Integration and acceptance of Natural User Interfaces for interactive learning environments International Journal of Child-Computer Interaction 2022 31 100381 10.1016/j.ijcci.2021.100381 Search in Google Scholar

Moore, B.A. and J. Urakami, The impact of the physical and social embodiment of voice user interfaces on user distraction. International Journal of Human-Computer Studies, 2022. 161: p. 102784. MooreB.A. UrakamiJ. The impact of the physical and social embodiment of voice user interfaces on user distraction International Journal of Human-Computer Studies 2022 161 102784 10.1016/j.ijhcs.2022.102784 Search in Google Scholar

Goodman-Deane, J., et al., A comparison of methods currently used in inclusive design. Applied Ergonomics, 2014. 45(4): p. 886–894. Goodman-DeaneJ. A comparison of methods currently used in inclusive design Applied Ergonomics 2014 45 4 886 894 10.1016/j.apergo.2013.11.00524456896 Search in Google Scholar

Ielegems, E., et al., ’Light up for all’ - Building Knowledge on Universal Design Through Direct User Contact in Design Workshops. Stud Health Technol Inform, 2021. 282: p. 102–119. IelegemsE. ’Light up for all’ - Building Knowledge on Universal Design Through Direct User Contact in Design Workshops Stud Health Technol Inform 2021 282 102 119 10.3233/SHTI21038834085962 Search in Google Scholar

Hurtienne, J., et al., Designing with Image Schemas: Resolving the Tension Between Innovation, Inclusion and Intuitive Use. Interacting with Computers, 2015. 27(3): p. 235–255. HurtienneJ. Designing with Image Schemas: Resolving the Tension Between Innovation, Inclusion and Intuitive Use Interacting with Computers 2015 27 3 235 255 10.1093/iwc/iwu049 Search in Google Scholar

Abascal, J. and C. Nicolle, Moving towards inclusive design guidelines for socially and ethically aware HCI. Interacting with Computers, 2005. 17(5): p. 484–505. AbascalJ. NicolleC. Moving towards inclusive design guidelines for socially and ethically aware HCI Interacting with Computers 2005 17 5 484 505 10.1016/j.intcom.2005.03.002 Search in Google Scholar

Langdon, P., U. Persad and P. John Clarkson, Developing a model of cognitive interaction for analytical inclusive design evaluation. Interacting with Computers, 2010. 22(6): p. 510–529. LangdonP. PersadU. John ClarksonP. Developing a model of cognitive interaction for analytical inclusive design evaluation Interacting with Computers 2010 22 6 510 529 10.1016/j.intcom.2010.08.008 Search in Google Scholar

Yang, J., et al., Incidence and Cost of Sexual Violence in Iowa. American Journal of Preventive Medicine, 2014. 47(2): p. 198–202. YangJ. Incidence and Cost of Sexual Violence in Iowa American Journal of Preventive Medicine 2014 47 2 198 202 10.1016/j.amepre.2014.04.00524930620 Search in Google Scholar

Mulfari, D., A. Celesti and M. Villari, A computer system architecture providing a user-friendly man machine interface for accessing assistive technology in cloud computing. The Journal of systems and software, 2015. 100: p. 129–138. MulfariD. CelestiA. VillariM. A computer system architecture providing a user-friendly man machine interface for accessing assistive technology in cloud computing The Journal of systems and software 2015 100 129 138 10.1016/j.jss.2014.10.035 Search in Google Scholar

Alomari, H.W., et al., A User Interface (UI) and User eXperience (UX) evaluation framework for cyberlearning environments in computer science and software engineering education. Heliyon, 2020. 6(5): p. e03917. AlomariH.W. A User Interface (UI) and User eXperience (UX) evaluation framework for cyberlearning environments in computer science and software engineering education Heliyon 2020 6 5 e03917 10.1016/j.heliyon.2020.e03917722666732426541 Search in Google Scholar

Yang, J. and R. Horie, An Improved Computer Interface Comprising a Recurrent Neural Network and a Natural User Interface. Procedia Computer Science, 2015. 60: p. 1386–1395. YangJ. HorieR. An Improved Computer Interface Comprising a Recurrent Neural Network and a Natural User Interface Procedia Computer Science 2015 60 1386 1395 10.1016/j.procs.2015.08.213 Search in Google Scholar

Yang, J., et al., Incidence and Cost of Sexual Violence in Iowa. American Journal of Preventive Medicine, 2014. 47(2): p. 198–202. YangJ. Incidence and Cost of Sexual Violence in Iowa American Journal of Preventive Medicine 2014 47 2 198 202 10.1016/j.amepre.2014.04.005 Search in Google Scholar

Kastner, P. and T. Dogan, Eddy, D: A toolkit for decoupled outdoor thermal comfort simulations in urban areas. Building and Environment, 2022. 212: p. 108639. KastnerP. DoganT. EddyD A toolkit for decoupled outdoor thermal comfort simulations in urban areas Building and Environment 2022 212 108639 10.1016/j.buildenv.2021.108639 Search in Google Scholar

Maskarenj, M., B. Deroisy and S. Altomonte, A new tool and workflow for the simulation of the non-image forming effects of light. Energy and Buildings, 2022. 262: p. 112012. MaskarenjM. DeroisyB. AltomonteS. A new tool and workflow for the simulation of the non-image forming effects of light Energy and Buildings 2022 262 112012 10.1016/j.enbuild.2022.112012 Search in Google Scholar

He, L., et al., Optimization-driven conceptual design of truss structures in a parametric modelling environment. Structures, 2022. 37: p. 469–482. HeL. Optimization-driven conceptual design of truss structures in a parametric modelling environment Structures 2022 37 469 482 10.1016/j.istruc.2021.12.048 Search in Google Scholar

Hsu, M., et al., An interactive geometry modeling and parametric design platform for isogeometric analysis. Computers & mathematics with applications (1987), 2015. 70(7): p. 1481–1500. HsuM. An interactive geometry modeling and parametric design platform for isogeometric analysis Computers & mathematics with applications (1987) 2015 70 7 1481 1500 10.1016/j.camwa.2015.04.002 Search in Google Scholar

Hughes, T.J.R., J.A. Evans and A. Reali, Finite element and NURBS approximations of eigenvalue, boundary-value, and initial-value problems. Computer Methods in Applied Mechanics and Engineering, 2014. 272: p. 290–320. HughesT.J.R. EvansJ.A. RealiA. Finite element and NURBS approximations of eigenvalue, boundary-value, and initial-value problems Computer Methods in Applied Mechanics and Engineering 2014 272 290 320 10.1016/j.cma.2013.11.012 Search in Google Scholar

Zhang, J., N. Liu and S. Wang, Generative design and performance optimization of residential buildings based on parametric algorithm. Energy and Buildings, 2021. 244: p. 111033. ZhangJ. LiuN. WangS. Generative design and performance optimization of residential buildings based on parametric algorithm Energy and Buildings 2021 244 111033 10.1016/j.enbuild.2021.111033 Search in Google Scholar

Herrema, A.J., et al., A framework for parametric design optimization using isogeometric analysis. Computer Methods in Applied Mechanics and Engineering, 2017. 316: p. 944–965. HerremaA.J. A framework for parametric design optimization using isogeometric analysis Computer Methods in Applied Mechanics and Engineering 2017 316 944 965 10.1016/j.cma.2016.10.048 Search in Google Scholar

Esfahani, S.K., et al., Optimizing the solar energy capture of residential roof design in the southern hemisphere through Evolutionary Algorithm. Energy and Built Environment, 2021. 2(4): p. 406–424. EsfahaniS.K. Optimizing the solar energy capture of residential roof design in the southern hemisphere through Evolutionary Algorithm Energy and Built Environment 2021 2 4 406 424 10.1016/j.enbenv.2020.09.004 Search in Google Scholar

Feng, K. and Z. Fan, A novel bidirectional LSTM network based on scale factor for atrial fibrillation signals classification. Biomedical Signal Processing and Control, 2022. 76: p. 103663. FengK. FanZ. A novel bidirectional LSTM network based on scale factor for atrial fibrillation signals classification Biomedical Signal Processing and Control 2022 76 103663 10.1016/j.bspc.2022.103663 Search in Google Scholar

Augustyniak, Ł., T. Kajdanowicz and P. Kazienko, Comprehensive analysis of aspect term extraction methods using various text embeddings. Computer Speech & Language, 2021. 69: p. 101217. AugustyniakŁ. KajdanowiczT. KazienkoP. Comprehensive analysis of aspect term extraction methods using various text embeddings Computer Speech & Language 2021 69 101217 10.1016/j.csl.2021.101217 Search in Google Scholar

Lindemann, B., et al., A survey on anomaly detection for technical systems using LSTM networks. Computers in Industry, 2021. 131: p. 103498. LindemannB. A survey on anomaly detection for technical systems using LSTM networks Computers in Industry 2021 131 103498 10.1016/j.compind.2021.103498 Search in Google Scholar

Wang, J., et al., Weighted IForest and siamese GRU on small sample anomaly detection in healthcare. Computer Methods and Programs in Biomedicine, 2022. 218: p. 106706. WangJ. Weighted IForest and siamese GRU on small sample anomaly detection in healthcare Computer Methods and Programs in Biomedicine 2022 218 106706 10.1016/j.cmpb.2022.10670635286872 Search in Google Scholar

ArunKumar, K.E., et al., Comparative analysis of Gated Recurrent Units (GRU), long Short-Term memory (LSTM) cells, autoregressive Integrated moving average (ARIMA), seasonal autoregressive Integrated moving average (SARIMA) for forecasting COVID-19 trends. Alexandria engineering journal, 2022. 61(10): p. 7585–7603. ArunKumarK.E. Comparative analysis of Gated Recurrent Units (GRU), long Short-Term memory (LSTM) cells, autoregressive Integrated moving average (ARIMA), seasonal autoregressive Integrated moving average (SARIMA) for forecasting COVID-19 trends Alexandria engineering journal 2022 61 10 7585 7603 10.1016/j.aej.2022.01.011 Search in Google Scholar

Busari, G.A. and D.H. Lim, Crude oil price prediction: A comparison between AdaBoost-LSTM and AdaBoost-GRU for improving forecasting performance. Computers & Chemical Engineering, 2021. 155: p. 107513. BusariG.A. LimD.H. Crude oil price prediction: A comparison between AdaBoost-LSTM and AdaBoost-GRU for improving forecasting performance Computers & Chemical Engineering 2021 155 107513 10.1016/j.compchemeng.2021.107513 Search in Google Scholar

Wang, J., et al., Weighted IForest and siamese GRU on small sample anomaly detection in healthcare. Computer Methods and Programs in Biomedicine, 2022. 218: p. 106706. WangJ. Weighted IForest and siamese GRU on small sample anomaly detection in healthcare Computer Methods and Programs in Biomedicine 2022 218 106706 10.1016/j.cmpb.2022.10670635286872 Search in Google Scholar

Shahid, F., A. Zameer and M. Muneeb, Predictions for COVID-19 with deep learning models of LSTM, GRU and Bi-LSTM. Chaos, Solitons & Fractals, 2020. 140: p. 110212. ShahidF. ZameerA. MuneebM. Predictions for COVID-19 with deep learning models of LSTM, GRU and Bi-LSTM Chaos, Solitons & Fractals 2020 140 110212 10.1016/j.chaos.2020.110212743754232839642 Search in Google Scholar

Ćalasan, M., S.H.E. Abdel Aleem and A.F. Zobaa, On the root mean square error (RMSE) calculation for parameter estimation of photovoltaic models: A novel exact analytical solution based on Lambert W function. Energy Conversion and Management, 2020. 210: p. 112716. ĆalasanM. Abdel AleemS.H.E. ZobaaA.F. On the root mean square error (RMSE) calculation for parameter estimation of photovoltaic models: A novel exact analytical solution based on Lambert W function Energy Conversion and Management 2020 210 112716 10.1016/j.enconman.2020.112716 Search in Google Scholar

Mohammadpourfard, M., et al., Cyber-Resilient Smart Cities: Detection of Malicious Attacks in Smart Grids. Sustainable Cities and Society, 2021. 75: p. 103116. MohammadpourfardM. Cyber-Resilient Smart Cities: Detection of Malicious Attacks in Smart Grids Sustainable Cities and Society 2021 75 103116 10.1016/j.scs.2021.103116 Search in Google Scholar

Articoli consigliati da Trend MD

Pianifica la tua conferenza remota con Sciendo