PRODUCTION LOGISTICS Predicting Slovak University of Technology Faculty of Material Science and Technology in Trnava.

PRODUCTIONLOGISTICS

Predicting

Slovak University of TechnologyFaculty of Material Science and Technology in Trnava

Predicting necessity Prediction helps the managers to reduce

uncertainty crating particulars plans. Predicting process estimates the products

requirements or these which will make in the future. It also guess the capacities, investments, profit, raw materials needs, energies, materials, staff, prices development, inflation etc. The centre of business demand remains the estimate of products’ sale.

Predicting Characteristic

Predicting methods suppose the system existence to which the prognosis is made for,

The predictions are only in abnormal case correct (mostly they are loose), the real results are different then the predicting ones,

Predicting Characteristic

Predicting is more narrow for products groups because the mistakes or particulars forecasts are compensating one another. Sources, raw materials, energies, staff are not mostly connected to actual products,

The forecast accuracy is falling down with length of its time level.

The Order of Predicting Process

1. Setting the prediction goal, date we are going to need it and sources necessary for predicting,

2. Defining the prediction time level,

3. Choosing the predicting methodology,

4. Collecting, analysis suitable information and their preparation. Hypotheses definition.

5. Monitoring and verification the prediction that is checking the prognosis was executing asked way. If not, the steps 2 till 4 will change.

Classification of predicting methods

2 ways: - qualitative,

- quantitative. Qualitative methods are based on subjective

customers’ information, vendors, managers, experts from which the prediction is made. Using: to prepare the prediction in short time or when we do not have enough information for using the qualitative methods.

Classification of predicting methods

Quantitative methods are based on analysis and preparing the historical data and their exploration to prediction season. They also find the causal relations of data time line used for prognosis calculation.

Quantitative Predicting Methods

Method based on the principle, the develop and relation between the values in the past will continue in the future. We can identify its last manner.Time line is time limited data sequencing, observations was obtaining in periodical (equal) time interval (hour, day, month, year etc.) e. g amount of requirements, prices, downtimes etc.Time level – it assigns how many intervals it is necessary to predict the quantity value forwards into the future.

Time aspect:

Strategic prognosis – is made for 3 till 5 intervals (years, months, weeks).It is composed for a consideration:

• Capacity rate on market (which firms are establishing or vanishing, how is their rate on market, how is the position of own business etc.),

• Outputs of consequential sectors,

Time aspect:

• The requirements to new products and the necessity of new investments,

• The crises, booms in particular productions group,

• Development of products, raw materials prices etc.

Time aspect:

Tactic prognosis displaces long time planning. It is mostly made for one time unit (year, month).It is needful for long cycle inputs orders and for capacity of preparation next production (machine capacity, staff, energies, costs, finance, profit etc.). It is input for lower levels of planning and production scheduling as well.

Fundamental Models of Time Line Manner

Constant model determines the values of the time line when in long time perion the data moves in narrow value interval disconnected by iregulavity, that is irregular, non repetitive, short time massive cahnge of some values of the time line or by value dispersion – caaualness.

Fundamental Models of Time Line Manner

Trend model – is characterized by permanent quality change (trend) of historical data time line. There are for example population curve, population incomes, volume of industrial production etc.

Fundamental Models of Time Line Manner

Cyclic model is characterized by periodical repetition of some tenor that is by some time there are repeating the value sequences of the time lines.

Fundamental Models of Time Line Manner

Season model is characterized by change of values in some term season fixed to weather (summer – holidays, the consumption is moving to another regions, winter – energy consumption, autumn – rising need of cannery tins) etc.

Fundamental Models of Time Line Manner

Combined models – consist of trend and cyclic models or trend and season ones.

Prediction Model

YN+1 = f(XN-i, T, tr, S, PC, Z, OPN+i, KZN+i)

when: YN+1 – the value of predicted parameter v season N+1, T – prediction time level, tr – trend, S – season, PC – cycle period, Z - casualness (fail), OPN+i – order in advance for season N+i, KZN+i – capacity changes in season N+i.

Characteristics for particulars types of models

K-model: YN+1 = f(XN-i, T, Z, OPN+i, KZN+i),

T-model: YN+1 = f(XN-i, T, tr, Z, OP N+i, KZN+i),

C-model: YN+1 = f(XN-i, T, PC, Z, OP N+i, KZN+i)

S-model: YN+1 = f(XN-i, T, S, Z, OP N+i, KZN+i),

ST-model: YN+1 = f(XN-i, T, tr, S, Z, OP N+i, KZN+i).

Applicability of particulars methods types for particulars kinds of models

K-model T-model C-model S-model

Arithmetical average

Balance models The same models as T one but there is applying the cycle period

The same models as K and T ones but there is applying

season index

Running average

Harmonic balance

Exponential balancing

Exponential balancing

Linear regression

Qualitative Predicting Methods

Qualitative methods comes out forecast,

experiences, practice and other form of

qualitative scoring information by people.

Qualitative Predicting Methods

There are using when:

Missing the historical data sequence (differ or it is anew product),

Change or political, economic situation (Slovakia after 1989), the conditions are changing on market and also in system.

The effort to create fast prediction.

Sorts of Qualitative Methods

• Estimation of vendors (market power),• Grouping survey,• Market analysis,• Delphi method.

Estimation of Vendors

This method is focused on periodic personal estimation future requirement by vendors.Advantages:

The vendors are such a group which can to forecast with high probability how product or service and in which amount will customer buy in nearest future.

If the company covers the market of large area, the prognosis will be divided into the region (segments). These information will be suitable for suppliers, distributors, vendors. Then we will get the prognosis as a sum of prognosis of cities, regions, areas, countries.

Estimation of Vendors

Disadvantages: Individual estimation of vendors (optimistic,

pessimistic), The vendors often do not differ what the

customer wants (or wish) and needs (is necessary for him) or just inquire about the foods.

If the company uses the prognoses of particular vendors than these have the interest that the store, office will have the perspective, they will have the job and this fact charges the estimation.

Grouping Survey

This method is based on knowledge, experience the specialists working in specific field (managers, finance experts, technical and technological specialists) and working out the prognosis together.

It mostly uses comparative approach and analogy of similar products sale, introducing new product or radical market change, political situation etc.

Market Analysis

Market Analysis is systematic approach of creating and hypothesis testing about market. The information is receiving various way. This methods is used introducing new product into the production. It consists of following steps:

1. Questionnaire design containing the questions about economic and demographic information about every respondent (person, company, institution) and the interview is made with them. Next part is about the questions including the interest about product putting on the market.

2. The choice of the communication way – telephone, fax, internet, mail, personal contact..

Market Analysis

1. The choice or representative respondents sample. I should be random sampling from next potential customers.

2. Data processing by prognosis after collecting questionnaires,

3. Creating the prognosis based on the build up and adapted data.

Delphi MethodIt is based on achieving the agreement between coordination group of specialists (Delphi committee) and anonym group of experts (they are not discussing between one and other about prognosis solution).Delphi committee figures out the communication way and formulate the questions on which ask the answer the experts. The questions must be refill the goal, facts on whit there is necessary to focus. These questions are sent to particular experts, they process the questions, to justify by facts, to support by arguments and scientific conclusions. Delphi committee will work out the attitudes and replies the experts then make the summarization and try to make the unite opinion – consensus (the first variant of prognosis). They will again sent it to experts and they will declare to design prognosis. This process has been repeating until the converge the attitudes will achieved. If the unit opinion will not reach then the experts can be changed or the process starts again with new experts group.

Prediction Errors

The error is the reason that the prognosis methods do not respect changing market conditions. The information’ absence for prognosis preparation (new production, step changes) minimize the chance to create the prediction with higher accuracy.

Prediction Errors is odds between requirements prognosis for season t and real requirements of season t.Et = Dt -YtWhere: Et – prognosis mistake in period t,

Dt – actual requirements for period t, Yt – prognosis for period t.

Types of Errors

Cumulated prognosis errors :

Quadratic deviations prognosis errors :

Standard deviations prognosis errors :

Average absolute deviations prognosis errors :

Average absolute percents of errors:

n

tEtCFE1

nEtMSEn

t/)(

1

2

)/)((1

2 nEtn

t

nEtMADn

t/)(

1

nDtEtMAPEn

t/)/)((

1

Types of Errors

MSE, MAD, calculated the variance the prediction error. If they are low, it do not means that the prediction is correct (the errors can compensate one another).

Average quadratic error underlines the extreme errors, positive and negative as well.

MAPE expresses the ratio of errors to real requirements.There is necessary to monitor the process of prediction, that is it is carried out right and if the right method was. Besides it, it is suitable to check out the error of prognosis moves to set interval. If the right method was used and the prognosis is calculated the right then the value of CFE should move about zero.Process signal:

n

tMADYtDtMADCFE

1/))((/