Descriptive Analytics
- • The primary stage of analytics in any decision-making scenario is descriptive analytics [33]. It provides solutions and explanations to the question of “What has happened” and apprehends and recapitulates the source data to a human-understandable format.
- • It engages with particulars of what took place in the past, what is happening currently, and why, illustrated in Figure 2.4. Also, it provides a clear and universal source of details throughout the supply chain. Descriptive Analytics identifies the possibilities and opportunities in the Supply Chain.
- • They are beneficial to exemplify the overall stock, the average amount consumed per user, and changes in sales per year. Some examples are intimation that furnishes information regarding past actions about the organisation’s manufacturing and presentation, economics and finances, operation and sales, and end-users.
- • Techniques and approaches such as Regression, Modelling and Visualisation, and OLAP (online analytical processing) are used in descriptive analytics
- • Descriptive Analytics functions can be divided into five categories:
- • Determining organisation goals and metrics: Performance is evaluated with respect to the goals and that makes determining the critical metrics very important. Enhancing the returns, lowering the expenses, increasing efficiency are some of the common goals.
- • Identification of the required data: There exist several sources from which the data can be extracted, such as desktops, warehouses, manual records. Understanding the requirements ensures proper planning to acquire the data and accurate extraction of the data [34].
- • Data Extraction and Preparation: This step although consumes the maximum quantity of time duration is essential to ensure high precision. Removal of duplicate entries, transforming data to a standard format, and cleansing are a few steps involved in data preparation.
- • Data Analysis and Processing: With the intention of identifying patterns and calculating performances, analysis models are created. Performance is evaluated for the initially mentioned goal by comparing it with past results [35]. R Programming and Python, open-source tools, are used to perform analysis.
FIGURE 2.4 Descriptive analytics.
• Data Presentation: Visual methods such as charts and graphs are commonly used to present the results obtained from the data processing and analysis. Data visualisation plays a vital role in this scenario. Several visualisation tools such as Business Intelligence provide the ability to present the data in a visual format.
Predictive Analytics
- • Predictive Analytics as shown in Figure 2.5 engages with the possibilities of what could happen. It strives to make accurate predictions about the future and explore the reasons. It assists an enterprise in understanding the probable results in the future and its implications on the business [36].
- • It can be applied during the business, from speculating the user behaviour and understanding designs to recognise inclinations in marketing and sales [37].
- • Quantitative and qualitative methodologies are both used to process and analyse the past and present data to make predictions. Its goal is to project what is the possibility of events occurring in the future and the reason behind its occurrence [38].
- • Predictive Analytics is associated with some techniques [39] and algorithms such as:
- 1. some methods used for estimating the sales in the supply chain are Advanced Forecasting and Time-series method
- 2. K-NN (Nearest Neighbour), Naive Bayes (NB), Discriminant Analysis are known Statistical algorithms being used for prediction
- 3. for the hierarchical consecutive structure, Random Forests and Decision trees are used
- 4. Identical or similar items in the humongous gathered data are grouped together accordingly using Clustering algorithms and Pattern-mining algorithms
- • Understanding predictive analysis by developing a predictive model using regression analysis [40].
- • Characterising the functioning of a random variable using a cluster of data mining or modelling approach with one or more numeric variables is known as regression.
- • The straight-line technique, linear regression, is used to determine the relationship between the predictor and response variable.
FIGURE 2.5 Predictive analytics.
• Straight-line represented by Equation 2.1
- - Regression is used to interpret the value of the variable у through finding the suitable values for the parameters m and c, using the value of A' as the foundation that is known.
- - Here, the dependent variable is у and the other variables become independent variables or predictor variables.
- • This methodology is preferred to learn the values of parameters associated with a function that has the capability of leading the function to its best-fit
- - In Equation 2.2, the value of Y, which is a continuous target, is to be estimated or predicted using regression function F.
- - (01, 02.. .On) is a set of parameters and the error is denoted by e
- • The result obtained or the output of the analysis is represented by R, known as the coefficient of determination. The value of R varies from Oto 1.
- - When R=0, the independent parameter is not useful in predicting the dependent parameter.
- - When 0>R< 1, it indicates that the range to which the depending parameter can be determined or predictable.
- - Example: When R=0.60, it specifies that 60% of the variance in the parameter у can be predictable from the parameter a.
Prescriptive Analytics
- • It is, as seen in Figure 2.6, concerned with what should occur in the future and what steps to be taken at the present to influence its occurrence - owning various judgements on the grounds of descriptive and predictive analytics, simulation, or mathematical optimisation, mainly building the knack of making decisions with several perspectives in mind [41].
- • What and when aspects are being dealt by descriptive and predictive analytics, while Prescriptive Analytics contemplates on the reason behind its occurrence (“why it occurred”) [42].
- • Data and information are collected continuously to trace back the events that provide the decision-makers with the opportunity to increase their prediction accuracy to make better choices.
- • This model is associated with optimisation and simulation. Its main aim is to enhance business performance by unravelling the reason for the occurrence of certain events.
FIGURE 2.6 Perspective analytics.
- • Although this analytic technology is comparatively complicated to apply [43]. But when applied appropriately, they have the strongest influence on how organisations make judgement calls and decisions.
- • The perspective analytic method uses the below mentioned two classes of algorithms
- • Decision trees
- - A tree-like graph structure or a model of decisions and their likely consequences that include the possibility of event outcomes, resource expense, and utility [44]. It is a technique to display an algorithm that encompasses only conditional control statements.
- - In the decision tree structure, each inner node marks a “test” on an attribute (such as, if a flipped coin lands on its head or tails), all the branches present represent the result of the trial or test, and leaf node represents the decisions taken after analysing. The route from the root to leaf represents classification rules.
- - For example: Critical Path Method (CPM) is used for project modelling, an algorithm for scheduling numerous activities and tasks related to projects. Another available method is known to be the Program Evaluation and Review Technique (PERT). Table 2.3 highlights the contrasting features of CPM and PERT. And having an understanding of contrasting characteristics of the two types of models, decision trees can be used to decide on which model can be used for implementation as shown in Figure 2.7.
- • Fuzzy Rule-Based System
- - The logic where more than true or false values are incorporated is known as fuzzy logic. This system comes into play when the situation cannot yield a direct true or false solution [45]. A continual range of truth values in the interval of [0, 1] is used rather than just the direct true or false values.
- - Fuzzy Rule-Based Systems are systems that embody an extension of traditional rule-based systems. Using fuzzy statements as the principal components of the rules permits gathering, apprehending, and managing the potential uncertainty of the represented knowledge. And, its structure is illustrated in Figure 2.8.
- - Inputs: A crisp numerical value. The inputs within the input subsets are combined wither with logical “AND” or logical “OR.”
TABLE 2.3
Difference in the Characteristic Between CPM and PERT
СРМ |
PERT |
Deterministic |
Probabilistic |
Uses historical or past data to make estimations |
Uses Probabilistic approach; hence there exists an opportunity for an activity to fail at any point. Therefore, the estimates are uncertain. |
Focuses on Expense and Time |
Focuses on planning |
Can be extended to small projects |
Suitable for R&D projects |
FIGURE 2.7 Decision tree for selecting a model.
- - Rules base: “IF....THEN...” statements are incorporated in the fuzzy rules [46]. Each rule is broken into two parts. The first part begins with an “IF” and terminates before the “THEN” is known as a predicate of combined inputs. The following consequent part that comes after “THEN” includes the subset of the output.
- - Applying the Implication Method: This phase is known to shape the consequence part.
FIGURE 2.8 Structure of fuzzy rule-based system.
- - Aggregating the outputs: It is necessary to combine all the rules, as the decisions are taken established on evaluation of all the conditional rules. In this step, output or result of each individual rule is integrated into one individual fuzzy set. The output obtained from the implication phase is used as the input to the aggregation process.
- - Defuzzification: This is the final phase where the obtained result in the fuzzy format is required to be converted to a crisp output using defuzzification techniques such as bisector, middle of maximum, that can be directly used.
- - Benefits of using Fuzzy Rule-Based Systems [47]:
- - Capable of providing an accurate solution to reason and rationalise with variability and uncertainty.
- - It is constructed from the experience of experts.
- - Liberal and tolerant towards ambiguous information and does not require historical data
- - Fuzzy rules can be developed easily using data from a survey without the need for in-depth pre-processing