Implementation of Lean and Six Sigma methodologies enable companies to boost their competitiveness and their efficiency. However, the adoption of these approaches is very much restricted in the Textile and Clothing sector in Morocco. In fact, despite all the advances in these methodologies and practical approaches, defining a rational implementation strategy such as the adequate chronology and the prediction of the expected success level are still a part of a fierce debate and an impediment for practitioners. The result is that only 11 companies out of 1,200 Moroccan clothing companies have successfully implemented Lean and Six Sigma. This article, based on an intelligent model, draws up a support tool to the clothing stakeholders, or otherwise aims to successfully integrate Lean and Six Sigma using Deep Learning. The neural network was trained for the prediction of success level rate and customizing of Lean and Six Sigma implementation chronology with the help of weights and maturity of a set of common critical success factors (CSFs). These CFSs were selected as input data. Then, the dataset have been used for training, testing, and validating the neural network model. To evaluate the trained network, 25% of the data have been used and a tuning hyperparameter process has been designed to reinforce the model performance. For the performance indices such as Categorical Cross Entropy (CCE), the defined loss function, accuracy, and precision have been evaluated and optimized. The developed model can then define the adequate chronology and predict success level with an accuracy of 97%. The trained neural network was then applied to a clothing company as a guide to the success of its continuous improvement project.
Lean Six Sigma is a combination of two powerful continuous improvement methodologies: Lean Management and Six Sigma. Lean philosophy focuses on reducing wastes and lead time through the operational process , while Six Sigma removes defects and increases quality. Each of these methodologies has become popularized individually by its successful implementation at pioneer companies such as TOYOTA and General Electric . Further, the integration of the two approaches can achieve better results than what either approach could not achieve alone  and provides a more significant impact to the organizational culture and the bottom line [4, 5].
Moreover, the choice of the implementation chronology or sequencing: (Lean followed by Six Sigma, Six Sigma followed by Lean, or both Lean and Six Sigma simultaneously) could have a direct effect on the implementation success level achieved .
Thus, the question of proper and fruitful sequencing causes a debate in the literature: Some authors insist on deploying Lean and Six Sigma simultaneously to guarantee optimal results . However, others argue that the application of both simultaneously may not give good results and should therefore be used separately: they advocate that Six Sigma should be used first, to increase process efficiency, and then Lean can be used to increase system efficiency . A third view suggests that it is more useful to implement Lean first for process sorting and waste reduction before using the complexity of Six Sigma to process and control the steps in each process . Then, a successful deployment of Lean Six Sigma depends on insightful implementing chronology.
Hence, this success is also based on many factors, identified as critical success factors (CSFs), has taken on a broader scope through significant studies [6, 10,11,12,13] and has been well expanded by establishing numerous CSF needed for fruitful implementation of Lean, Six Sigma, and Lean Six Sigma .
This research paper aims to outline the choice and sequence to implement successfully either Lean or Six Sigma based on the common CSFs features. This is the first study that has used a neural network model to predict the potential success level and indicate the required chronology to implement Lean manufacturing and Six Sigma based on their CSFs. The developed model will be applied within the clothing sector.
Thus, the purpose of this study is firstly to retrieve the required data and link them (Figure 1).
The data will be compiled from:
CSFs weight, which is the amount of importance allocated to each CSF: in this purpose, Lean and Six Sigma common CFSs's weights will be handled independently.
CSFs maturity, which corresponds to the level of their deployment. It is referring to the depth of the embodiment of each factor in the company, through its extension into the organization .
The implementation chronology is chosen and applied by the studied companies. Then we set three possibilities: Lean followed by Six Sigma; Six Sigma followed by Lean; and both Lean and Six Sigma simultaneously.
The level of success is achieved from the continuous improvement approaches’ deployment. It refers to the achievement of greater business performance through higher process efficiency and by elimination of manufacturing wastes .
Obviously, given the complexity of the problem and the multitude of variables, the answer will be provided by a trained neural network model (Figure 1).
Moreover, we had chosen the clothing and textile sector to apply this model, since today in Morocco, this sector has a tendency to raise its competitiveness by tightening the prices, and there is no tolerance for any waste or scraps that could affect the targeted profit margin.
This paper is drawn up as follows: Session 2 displays the common CSFs of Lean and Six Sigma extracted from the literature. Session 3 describes the three steps of the methodology: In the first step, the CSFs were confirmed by the interviewed experts who had also noted, based on the Likert scale, the maturity and weight of each CSF within their companies. Also, they had indicated the chosen chronology and the achieved success level of Lean and Six Sigma. In the second step, a test of reliability was employed for all of the data, and also the Delphi method was employed to assure the validity of the CSFs’ required maturity. In the third step, the neural network is employed as a new method for predicting the scores of success and determining the appropriate chronology as output. Then, we trained our neural network based on the reliable database of CSFs’ weight and on consented maturity as inputs (Figure 1). Session 4 and 5 exemplify the graphical and numerical results and discussion. Session 6 illustrates, within a clothing company, a recommended method of building on the developed model to prepare for the Lean and Six Sigma implementation phase.
2 Literature review
The literature revealed a set of factors that function as a prerequisite for successful implementation of Lean and Six Sigma . These factors are frequently reported as CSFs [6, 18]. Thus, many companies are adopting the CSF strategy to guarantee the success and sustenance of Lean manufacturing, Six Sigma, and Lean Six Sigma implementation and furthermore to ensure the greatest profitability .
Then, defining the CSFs helps companies, notably clothing companies, to achieve a high success rate for Lean and Six Sigma projects , and facilitates the choice between the two approaches, encourages their simultaneous integration, and helps to define the most suitable chronology that varies with each company context .
One of the main reflections made is that the CSFs are similar for all the change and improvement approaches, but only with slight variations . These factors have then became common CSFs for Lean and Six Sigma and are set out in Table 1.
|Common CSF||References of Lean||References of Six Sigma|
|CSF1: Management commitment||[12, 19]||[3, 17]|
|CSF2: Culture change||[3, 17]||[6, 20]|
|CSF3: Communication||||[22, 23]|
|CSF4: Training and education of employees||[16, 23, 24]||[12, 16, 23]|
|CSF5:Understanding of tools and techniques||[6, 15]||[6, 15, 16]|
|CSF6: Employee involvement||[2,3,4, 14]||[6, 12]|
|CSF7: Skills and expertise||||[3, 12]|
CSFs, critical success factors.
However, based on this overview where the common CSFs of Lean and Six Sigma have been set, all the experts and researches claimed that they are not all on the same level of importance for the same approaches [6, 14].
Furthermore, the required maturity level of each CSFs deployment is not on the basis of equality.
Then, success of the Lean and/or Six Sigma implementation relies on measurement of companies’ readiness made through a set of CSFs, as mentioned above, by assigning each of them the right weight and the required maturity. Moreover, handling of the CSFs to prepare them for the pre-implementation phase is not only conditioning them for the sequencing of Lean Six Sigma implementation but also the anticipated success level.
For this purpose, we had to prepare our data to ensure the suitability of all the above variables to our model. This will be accomplished in the next section.
The Machine Learning Algorithms and neural networks are good alternatives to this kind of problem that require data insight and training . Also, unlike many heuristic algorithms, deep learning does not have any calculation rules and can be trained by data to identify strong correlations between CSFs’ weight and maturity (Inputs) and Chronology and success level (outputs).
However, the greatest challenge of this approach is the lack of large datasets of actual Lean and Six Sigma implementation scenarios.
Then, to overcome this impediment, the first step of this work consists of building a strategy to get a comprehensive and reliable database. The two actions to accomplish this task are: collecting and processing, via the reliability test and Delphi method, the data to render them into appropriate input parameters for the neural network.
In summary, a combination of approaches has been employed in this paper to collect the data and evaluate them.
Then, as described in Figure 1, 21 inputs and 2 outputs were identified; among the input data, 14 are related to CSFs’ weight and 7 to the CSFs maturity. Among the output data one is related to chronology and one to the success level.
3.1 Data collecting
This step undertook a measurement of the applied weights and maturity to Lean and Six Sigma's CSFs within companies that have dealt with their implementation under different chronologies. It also covered companies that have failed to implement the two methodologies.
To participate in this study, the practitioners should have managed at least one Lean and Six Sigma project and should be Lean Six Sigma certified.
Data were collected via an online survey.
The implementation success level and CSFs weights measurements were conducted by the five-point Likert scale. The respondents were asked to choose from the following indicators: 1 = failure; 2 = slightly successful; 3 = successful; 4 = very successful; and 5 = extremely successful.
Also, they were invited to fill the weight of each CSFs from the perspective of each approach using the following scale: 1 = non important; 2 = slightly important; 3 = important; 4 = very important; and 5 = extremely important.
To identify the decided chronology within each company, every expert had to choose from three sequencing patterns that have been coded into numbers: (1 – Lean followed by Six Sigma, 2 – Six Sigma followed by Lean, and 3 – Lean and Six Sigma simultaneously).
However, for CSFs maturity assessment, the response choices for maturity range from 10% to 100%. The higher the percentage is, the more mature the CSFs is.
3.2 Data processing
It has been shown that, in addition to the hyperparameter tuning, a neural network model works well when supported by reliable or good quality data . Thus, to test the reliability of the collected data, Cronbach's alpha was used and it showed a reliable internal consistency: (0.814–0.858) for the CSFs’ weight and (0.944–0.953) for Lean and Six Sigma success level. This indicates good internal consistency of the data gathered (i.e., in the range of 0.814–0.953 with a value of >0.7). Thus, we concluded that the survey is reliable .
Also, given the complex and subjective nature of assigning the required CSFs’ maturity for a successful integration of Lean and Six Sigma, the Delphi method is an appropriate tool for determining consensus among Moroccan industry experts.
In fact, consensus building is a key component of any evaluation process. In addition, the strengths of the Delphi method are to bring stakeholders and experts with opposing views to facilitate consensus as well as to highlight the divergence in the opinions . Thus, it is an accurate tool as it can cover a multiplicity of issues and offer guidance for the intended conventional consent.
In fact, participating in the Delphi process and being a stakeholder of defining an agreement on the maturity of Lean Six Sigma LSFs between key players in the field of implementing continuous improvement approaches can be a highly motivating experience for participants.
Also, the greatest strength of the Delphi method is that it integrates education and consensus building into the multistage process of data collection (Figure 2); thus we will proceed via a multistage process involving the initial collection of opinions; followed by the analysis of the data; then designing a new interview including, anonymously, their answers and those of others, which were read aloud; and then the experts were invited to express their opinion on the heard answers. The third step consists of asking the experts to assess their agreement with the results obtained during the first and second steps.
The purpose of the study was to measure the consensus degree among CSFs’ maturity and assessing its shift over different stages.
However, measure of the consensus varies from study to study . Frequency distributions are used to assess agreement , and also the criterion of at least 51% response category . Another study use the yes–no response category, and the criterion for agreement was 67% of participants giving an identical response .
Others had used interquartile deviation (IQD) to reach the consensus. The interquartile range is the difference in absolute value, between the 75th and 25th percentiles , with smaller difference indicating a higher degree of consensus. In this article we will use the interquartile method.
An IQD of 1.00 or less was considered as an indicator of consensus , while  we considered a change of >1 IQD point from stage to stage as a consensus indicator.
Apparently, there is no consensus on how to apply the IQD as a data analysis indicator for the Delphi process. So, the strategy to use and interpret the IQD is as follows:
CSFs with IQD = 0.00 were considered to reflect consensus and were not included in the second stage.
The CSFs with IQD = 1.00 also were omitted in the second stage because there was a high degree of agreement among respondents. But other CSFs with IQD = 1.00 were included in the third-stage interview guide because there was considerable variability in the distribution of responses among these items. On the maturity scale, the variability is detected when the absolute value of the difference between the percentage of respondents who were “generally mature” (i.e., Totally Mature, Very Mature, and Mature) and the percentage who were “generally immature” (i.e., Slightly Mature and not Mature) is >20%. (Table 2).
|CSFs||Maturity Level||%||IQD||| Generally positive - Generally negative ||
In summary, we define two conditions to demonstrate the consensus:
Absolute value of the difference between the percentage of respondents who were generally mature and the percentage who were generally immature ≥20%.
This selection method was allowed at the end of the third stage, on which CSFs the experts had the least agreement (CSF3: Communication and CSF4: Training and Education of employees).
3.3 The neural network modeling and training
The neural networks consist of a set of processing elements called “neurons.” It is an interconnected group of nodes that is used for complex relationship between inputs and outputs . The Figure 3 shows an example of a neural network architecture.
This neural network has a double hidden layer – R inputs (I) and S output (O). Each neuron computes the sum of the input weights in the presence of a bias, and passes the sum to the activation function to obtain the output.
The main challenge is choosing the right training algorithm. Furthermore, the design of neural networks is very complicated because many factors affect the training activities, such as the number of neurons in the hidden layer, the link between neurons and the layer, error function, and activation function .
Thus, the hyperparameter tuning is a crucial step in optimizing neural network results. It is a technique where the parameters of algorithm are chosen in such a way that the optimal solution can be obtained, which depends critically on hyperparameter settings .
Table 3 describes the hyperparameter combinations of our neural network learning algorithm. In our setup, there are four hyperparameters to tune (Number of hidden layers, Dropout, Learning rate, Optimizer), and 108 alternatives have been evaluated.
|Number of neurons||16||32||64||128|
In this paper, we will use the exact hyperparameter optimization algorithms, one of the most commonly used, which has been shown to give good empirical results for hyperparameter tuning.
4 Results and discussion
4.1 The proposed framework
Our algorithm, during the process, kept on adding layers until the test error ceased to show any improvement. We obtained, successively, 128 units and 16 units located in the first and second hidden layer of the algorithm, respectively, in which the Relu function applied weights to the inputs.
To avoid overfitting and increase the generalizing power, the best value defined for dropout in a hidden layer was 0, 1.
To train the neural network using different adaptive learning rate, Rmsprop was defined as optimizer.
Then, 0.001 was defined as learning rate and the network is allowed to adequately update its parameters and maintain the best weight control at the end of each batch. It also controls how quickly or slowly a neural network model learns our problem. Then, Rmsprop optimizer maintains and adapts this defined learning rate for each of the weights in the model.
The Relu is a linear function that will output the input directly if it is positive; otherwise, it will output zero. This activation function overcomes the vanishing gradient problem, allowing models to learn faster and perform better.
The Relu is a linear function that will output the input directly if it is positive; otherwise, it will output zero. This activation function overcomes the vanishing gradient problem, allowing models to learn faster and perform better.
For the multiclass classification, the softmax function is used as the activation function in the output layer of the neural network models that predict a multinomial probability distribution. We set the softmax activation to output one value for each node in the output layer.
Overall, the hyperparameters were tuned through a staged process. The tuning process of the best hyperparameter combination in terms of validation performance was conducted through four mains steps as shown in Figures 4 and 5.
Finally, assessing performance through accuracy allowed us to obtain conclusions on the efficiency of our network performance.
As mentioned in the previous session, we found that the best results were obtained from these hyperparameter combinations:
Number of Hidden layers: 2
Number of units in the first layer:128
Number of units in the first layer:16
Learning rate: 0.001
4.2 The neural network model evaluation
One of the main indications of the overfitting is increase in the validation error. In our multiclass classification case, some of the most commonly used criteria were chosen, namely Categorical Cross Entropy (CCE) and Accuracy, to further evaluate the performance of the network model.
Thus, our model uses the CCE to learn to give a high probability to the correct digit and a low probability to the other digits.
Furthermore, the Accuracy evaluates the precision of our model compared to the actual data points. The higher the Accuracy is, the more efficient is the model .
The best neural network is characterized by small CCE and high Accuracy.
Figures 6 and 7 compare and illustrate the CCE and the Accuracy evolutions as a function of the number of epoch. By reaching 100 of the epoch number, the CCE marks the lowest value (0.145) while the accuracy reaches the highest one (97%).
For instance, our model proposed 100 as the epoch number to get good results with a high accuracy of 97%.
Furthermore, three metrics, in addition to classification accuracy, are commonly required for the evaluation of neural network model performance based on a test dataset .
The metrics used are precision, recall, and specificity (Table 4):
Precision defines the percentage of samples with a certain predicted class label actually belonging to that class label: High precision relates to the low false positive rate. We have got 96. 3% precision, which is pretty good.
Recall defines the percentage of samples of a certain class, which were correctly predicted as belonging to that class: We achieved of 96. 8%, which is good for this model as it is above 50%.
Specificity shows how good the test is at predicting: The higher value of specificity would mean higher value of true negative and lower false positive rate. In our case, specificity score is 96. 9%.
In summary, the bias and the weights were chosen randomly at first; afterward, neural network will learn by itself through application of multiple iterations doing forward propagation while marking the highlighted metrics in this session.
4.3 Applying the trained neural network to a clothing company
The studied company SAN-STYLE is one of the oldest and most competitive Moroccan textile enterprises from the clothing branch.
SAN-STYLE is located in Casablanca, Morocco and is part of the Wholesale Sector Textile Industry. SAN-STYLE has a total of 150 employees across all of its locations and generates US$ 2.35 million in sales.
The company's operational activities include the creation, sourcing, and production of textile and clothing articles, especially tailoring, sewing, weaving, and clothing transformation. The company has already tried to implement Lean and Six Sigma; we will then compare the actual results with those predicted.
Then, in this session, we will apply the neural network model developed within the clothing company to define the appropriate chronology of implementing Lean and Six Sigma. Furthermore, the results obtained can be used to predict the success level at pre-implementation phase. Then, the model is going to help the textile and clothing industry managers to develop a customized integration strategy during their continuous journey of improvement.
Field data are collected from manufacturing site over a period of 3 months. The data collection exercise involves 21 input variables (CSFs weight and maturity) and 2 output variables (chronology and success level). The input will get into our trained neural network model and will be used for defining and predicting the output data within SAN-STYLE Company (Figure 8).
In general and for any other company, Figure 8 explains the steps involved in training and applying the proposed model.
According to the results, the predicted success level while implementing Lean and Six Sigma simultaneously, as also recommended by the neural network model, is 60% of the targeted aim of reducing wastes, lead time, variability of the manufacturing process, and improving the customer service and the added value.
Furthermore, comparing the result of the factual implementation strategy conducted randomly by the studied company and those suggested by the use of the trained neural network model, we revealed that they had achieved only 40% of the expected success level instead of 60% by implementing first Lean manufacturing and then the Six Sigma instead of their simultaneous implementation.
As found out, and according to the respective values of accuracy and CCE of (97%) and (0.145), the neural network is a good approach for minimizing the uncertainties in the continuous improvement projects and achieving the best result.
In fact, they have implemented and achieved the best results based on a trained model of data from different companies. The developed neural network gives to the company the ability to meet their CSFs and resources with better results.
Concretely, SAN-STYLE Company, based on its current resources in terms of their CSFs’ maturity level, and assigned weights, has the opportunity to adjust its implementation strategy to achieve 20% more than the current result.
In parallel, SAN-STYLE can conduct a continuous learning as another important advantage for the chronology and success level prediction, because training data is not limited and new cases are continuously encountered. For this reason, the use of neural network may provide the company a new approach with which to enhance constantly its implementing strategy performance while increasing the inconsistency of correlations between CSFs and the targeted performance.
To correctly manage the integration of the continuous improvement approaches of Lean and Six Sigma and at the same time guarantee the highest success level, a methodical strategy based on the assessment of their CSFs is recommended.
In fact, the lack of a common ground around the integration chronology, and moreover a lack of the common language to decide, often complicates the alignment of these two approaches that are significantly different but highly connected.
Then, the solution of this complex problem required an interdisciplinary study:
A state of art of Lean and Six Sigma common CSFs.
An assessment of their maturity and through interviews with experts.
A multistage process to get the expert and certified practitioners’ consensus about the required maturity to each CSFs.
A collection of the chronology adopted by companies that have effectively integrated Lean and Six Sigma.
A collection of a large amount of this data from industrial companies that have effectively integrated Lean or Six Sigma or Lean Six Sigma.
We developed a decision support tool based on these observations that, reliably and simultaneously, determines the chronology that is appropriate to each case and the fitted success level to expect.
As establishing a relationship between input (CSFs Lean weight, CSFs Six Sigma weight, and CSFs maturity) and output (Chronology and integration success level) variables seemed complex, the neural network was applied.
So, it was necessary to create a reliable database of training and validation data to apply this methodology.
Hence, we created data via an online survey and purged the database based on statistical analysis via the alpha-Cronbach and Delphi methodology.
These indicators’ results were:
Dataset alpha-Cronbach: (0.814–0.858) for the CSFs weight, and (0.944–0.953) for Lean and Six Sigma success level.
Delphi Consensus conditions: IQD ≤1 and the absolute value of the difference between the percentage of respondents who were generally mature and the percentage who were generally immature ≤20%.
This resulting refined database, constituted by 21 inputs and 2 inputs, was used to train the neural networks.
Several hyperparameters of the neural network were analyzed and the best combination was described.
Following the identified accuracy (97%), the loss function (0.145), the metrics of precision (96, 3%), Recall (96, 8%), and Specificity (96, 9%), the details of the model's performance evaluation were represented by Section 4 (Figures 6 and 7, and Table 4).
The good results and high degree of reliability emphasize the use of the neural network as an excellent alternative method to solve the complex problem of deciding on the adequate chronology of integrating the Lean and Six Sigma and to predict the success level based on the tuned CSFs weight and maturity.
In this paper, the neural network model application was conducted within a clothing company. We simplified the implementing strategy to enable adoption based on its CSFs, which give an optimal performance.
This work provides for clothing and textile companies an instrument that resolves the fierce debate about chronology with a common tool and a single language.
Additionally, this model is generic and can be applied in any sector for any company that aims to succeed in its Lean Six Sigma implementation journey.
And the model application would provide clothing managers with guidelines for a successful implementation of Lean principles and Six Sigma techniques, which allows optimizing time and resources.
Indeed, this case study in the clothing context enabled the understanding of a methodology that can be extended to any context and any condition.
 Larsson, J., Peterson, J., Mattila, H. (2012). The knit on demand supply chain. Autex Research Journal, 12(3), 67–75.10.2478/v10304-012-0013-9Search in Google Scholar
 O’Rourke, P. (2005). A multiple-case comparison of lean six sigma deployment and implementation strategies. In ASQ World Conference on Quality and Improvement Proceedings, 59, 581.Search in Google Scholar
 Antony, J. (2011). Six sigma vs lean: Some perspectives from leading academics and practitioners. International Journal of Productivity and Performance Management, 60(2), 185–190.10.1108/17410401111101494Search in Google Scholar
 Cudney, E. A., Kestle, R. (2018). Implementing Lean Six Sigma Throughout the Supply Chain: The Comprehensive and Transparent Case Study. Productivity Press (Routledge).10.1201/b10290Search in Google Scholar
 Salleh, N. M., Ne, P. (2019). Comparative study between lean six sigma and lean-agile for quality software requirement. International Journal of Advanced Computer Science and Applications, 10(12).10.14569/IJACSA.2019.0101230Search in Google Scholar
 Alhuraish, I., Robledo, C., Kobi, A. (2017). A comparative exploration of lean manufacturing and six sigma in terms of their critical success factors. Journal of Cleaner Production, 164, 325–337.10.1016/j.jclepro.2017.06.146Search in Google Scholar
 Salah, S., Rahim, A., Carretero, J. A. (2010). The integration of six sigma and lean management. International Journal of Lean Six Sigma, 1(3), 249–274.10.1108/20401461011075035Search in Google Scholar
 Crawford, R. (2004). Ammunition enterprise excellence ready for tomorrow?, USA Armor School Research Library.Search in Google Scholar
 Snee, R. D. (2004). Six–sigma: The evolution of 100 years of business improvement methodology. International Journal of Six Sigma and Competitive Advantage, 1(1), 4–20.10.1504/IJSSCA.2004.005274Search in Google Scholar
 Bakar, F. A., Subari, K., Daril, M. A. (2015). Critical success factors of lean six sigma deployment: A current review. International Journal of Lean Six Sigma, 6(4), 339–348.10.1108/IJLSS-04-2015-0011Search in Google Scholar
 Rosa, A. C. M., Broday, E. E. (2018). Comparative analysis between the industrial and service sectors: A literature review of the improvements obtained through the application of lean six sigma. International Journal for Quality Research, 12(1), 227.Search in Google Scholar
 Mishra, M. N. (2018). Identify critical success factors to implement integrated green and lean six sigma. International Journal of Lean Six Sigma.10.1108/IJLSS-07-2017-0076Search in Google Scholar
 Stankalla, R., Koval, O., Chromjakova, F. (2018). A review of critical success factors for the successful implementation of lean six sigma and six sigma in manufacturing small and medium sized enterprises. Quality Engineering, 30(3), 453–468.10.1080/08982112.2018.1448933Search in Google Scholar
 Elboq, R., Hlyal, M., El Alami, J. (2020). Lean manufacturing and six sigma critical success factors: A case study of the Moroccan Aeronautic Industry. International Journal of Supply Chain Management, 9(4), 2051–3771.Search in Google Scholar
 Cheng, C. Y., Chang, P. Y. (2012). Implementation of the lean six sigma framework in non-profit organisations: A case study. Total Quality Management and Business Excellence, 23(3–4), 431–447.10.1080/14783363.2012.663880Search in Google Scholar
 Ali, N. K., Choong, C. W., Jayaraman, K. (2016). Critical success factors of lean six sigma practices on business performance in Malaysia. International Journal of Productivity and Quality Management, 17(4), 456–473.10.1504/IJPQM.2016.075251Search in Google Scholar
 Maleyeff, J., Arnheiter, E. A., Venkateswaran, V. (2012). The continuing evolution of lean six sigma. The TQM Journal.10.1108/17542731211270106Search in Google Scholar
 Näslund, D. (2013). Lean and six sigma–critical success factors revisited. International Journal of Quality and Service Sciences, 5(1), 86–100.10.1108/17566691311316266Search in Google Scholar
 M. Bożek and A. Hamrol, “Analysis of efficiency of Lean Manufacturing and Six Sigma in a production enterprise,” Manag. Prod. Eng. Rev., vol. 3, no. 4, pp. 14–25, 2012.Search in Google Scholar
 Netland, T. H. (2016). Critical success factors for implementing lean production: The effect of contingencies. International Journal of Production Research, 54(8), 2433–2448.10.1080/00207543.2015.1096976Search in Google Scholar
 Antony, J., Manville, G., Greatbanks, R., Krishnasamy, R., Parker, D. W. (2012). Critical success factors for lean six sigma programmes: A view from middle management. International Journal of Quality and Reliability Management.10.1108/02656711211190846Search in Google Scholar
 Antony, J., Banuelas, R. B. (2002). Critical success factors for the successful implementation of six sigma projects in organizations. The TQM Magazine, 14(2), 92–99.10.1108/09544780210416702Search in Google Scholar
 Ho, Y. C., Chang, O. C., Wang, W. B. (2008). An empirical study of key success factors for six sigma green belt projects at an Asian MRO company. Journal of Air Transport Management, 14(5), 263–269.10.1016/j.jairtraman.2008.05.002Search in Google Scholar
 Ortiz, C. A. (2012). The psychology of lean improvements: Why organizations must overcome resistance and change the culture. Productivity Press (Routledge).10.1201/b11898Search in Google Scholar
 Sobral, M. C., de Sousa Jabbour, A. B., Chiappetta Jabbour, C. J. (2013). Green benefits from adopting lean manufacturing: A case study from the automotive sector. Environmental Quality Management, 22(3), 65–72.10.1002/tqem.21336Search in Google Scholar
 D’Amico, B., Myers, R. J., Sykes, J., Voss, E., Cousins-Jenvey, B., et al. (2019). Machine learning for sustainable structures: A call for data. In Structures, 19, 1–4.10.1016/j.istruc.2018.11.013Search in Google Scholar
 Farooq, A., Irshad, F., Azeemi, R., Iqbal, N. (2020). Prognosticating the shade change after softener application using artificial neural networks. Autex Research Journal, 1, 79–84.10.2478/aut-2020-0019Search in Google Scholar
 Gliem, J. A., Gliem, R. R. (2003). Calculating, interpreting, and reporting Cronbach's alpha reliability coefficient for Likert-type scales. Midwest Research-to-Practice Conference in Adult, Continuing, and Community Education.Search in Google Scholar
 Rayens, M. K., Hahn, E. J. (2000). Building consensus using the policy Delphi method. Policy, Politics, & Nursing Practice, 1(4), 308–315.10.1177/152715440000100409Search in Google Scholar
 McKenna, H. P. (1994). The delphi technique: A worthwhile research approach for nursing? Journal of Advanced Nursing, 19(6), 1221–1225.10.1111/j.1365-2648.1994.tb01207.xSearch in Google Scholar PubMed
 McKenna, H. P. (1989). The selection by ward managers of an appropriate nursing model for long-stay psychiatric patient care. Journal of Advanced Nursing, 14(9), 762–775.10.1111/j.1365-2648.1989.tb01641.xSearch in Google Scholar PubMed
 Alexandrov, A. V., Pullicino, P. M., Meslin, E. M., Norris, J. W. (1996). Agreement on disease-specific criteria for do-not-resuscitate orders in acute stroke. Stroke, 27(2), 232–237.10.1161/01.STR.27.2.232Search in Google Scholar PubMed
 Raskin, M. S. (1994). The delphi study in field instruction revisited: Expert consensus on issues and research priorities. Journal of Social Work Education, 30(1), 75–89.10.1080/10437797.1994.10672215Search in Google Scholar
 Spinelli, T. (1983). The delphi decision-making process. The Journal of Psychology, 113(1), 73–80.10.1080/00223980.1983.9923559Search in Google Scholar
 Ghith, A., Hamdi, T., Fayala, F. (2015). Prediction of drape coefficient by artificial neural network. Autex Research Journal, 15(4), 266–274.10.1515/aut-2015-0045Search in Google Scholar
 Khan, Y., Shafiq, S., Naeem, A., Hussain, S., Ahmed, S., et al. (2019). Customers churn prediction using artificial neural networks (ANN) in telecom industry. Editorial Preface From the Desk of Managing Editor, 10, (9).10.14569/IJACSA.2019.0100918Search in Google Scholar
 Khot, A., Dave, M. (2021). Position falsification misbehavior detection in VANETs. In Mobile Radio Communications and 5G Networks, Springer (Singapore), 487–499.10.1007/978-981-15-7130-5_39Search in Google Scholar
 Elhami, S., Razfar, M. R. (2017). Analytical and experimental study on the integration of ultrasonically vibrated tool into the micro electro-chemical discharge drilling. Precision Engineering, 47, 424–433.10.1016/j.precisioneng.2016.09.015Search in Google Scholar
© 2021 Raja Elboq et al., published by De Gruyter
This work is licensed under the Creative Commons Attribution 4.0 International License.