Observation and measurement

Many computer systems record the information of various objects in the real world. Such information is usually represented by records, attributes, objects, and other forms in computer systems. The most typical way is to record a certain item of information as an attribute of an object. For example, a person weighing 70 kilograms is recorded as a weight (weight) attribute of the "person" class, the value is 70. This chapter describes the shortcomings of this method and provides some more reasonable solutions.

The model in this chapter comes from projects related to the medical field, so many examples are used in this field.

The pattern diagrams in this chapter are all re-drawn by the author in the general UML format.

Quantity 1.1 (Quantity)

When using the method mentioned above to record data, the most common disadvantage is that a simple number is not enough to represent its meaning. It is 70 kilograms, 70 lbs, or something else? We need the exact unit. Can I create associations between attributes and units? Yes, but the system will have a complex association, which increases the complexity of the system.

If a quantity class is used for expression, the meaning will be simpler. As shown in, the quantity class includes an amount attribute, a record value, an units attribute, and a record unit. It also supports general operations.

1.2 conversion rate (Conversion
Ratio)

This mode mainly solves the problem of conversion between different units. By converting different units and the ratio between them, you can achieve unit conversion of various fixed ratios, this is also very useful in the system.

However, it also has limitations. If the conversion ratio is not fixed, additional computing functions may be required. Generally, this mode is sufficient.

1.3 Compound
Unit)

The unit can be composite or non-decomposed (Basic Unit). How can we express the composite unit? See the following two models:

In this model, compound unit records the basic unit and their power through unit reference ). This is a relatively direct model.

The following model is more compact due to the use of bags:

The two models have little difference, but one uses the bag, and the other uses the unit reference. Do you want to introduce a composite unit when the model is used, it depends entirely on the needs of customers and applications.

1.4 mesurement)

When a complex system contains thousands of measurement activities and the measurement data needs to be recorded, relying solely on the quantity model is not enough. If the measurement data is still processed as an attribute, the system may be filled with disorganized attributes, resulting in some very complex interfaces ). The solution here treats various measurement items (such as the height, weight, blood pressure, and blood sugar concentration required by hospitals) as objects, the object type is introduced as "phenomenon type", and the complexity of the problem is transferred to various phenomenon types and mesurement instances.

As shown in the model, the mesurement instance already includes the phenomenon type, measurement object (person), and result (Quantity ). According to the separation of the Knowledge layer and the operation layer mentioned in the previous chapter, we can divide the phenomenon type into the knowledge layer, because these objects represent thousands of measurement items, this part of information remains unchanged.

You can also associate the unit information with the type (phenomenon type), while the measurement instance only saves simple values. However, to support multiple units for the type (phenomenon type, we are still inclined to adopt the model.

1.5 observation mode)

In addition to recording the specific values of many measurement items, such as height, weight, and blood glucose concentration, a measurement item only needs to record the type of the item, such as blood type, there are four common types: A, B, O, and AB. For example, the shape can be recorded as obese, normal, and thin, therefore, it is necessary to add categories in the above model to form the observation mode:

Here, category is best associated with the symptom type to reduce ambiguity and complexity. In the following figure, category is renamed to phenomenon and moved to the knowledge layer. phenomenon defines some phenomenon.
A group of possible values of type.

For observation, it should also be mentioned that in the actual diagnosis process, the doctor needs to infer certain measurement activities based on certain phenomena, these measurement activities provide evidence for other measurement records, so there may be the following relationships:

1.6 improvement on Observation

In the above figure, a possible case is missing, that is, for some record items, the result may be a simple "presence) "or" No/NO (absense) ", for example, for various diseases, a person may have diabetes, maybe not. This is why the category observation has two subclasses: absence and presence.

Here, a parent class observation concept is added for phenomenon, so that they do not need to be associated with phenomenon type when dealing with such diseases.

Also, we can see that a supertype self-Association is added to the observation concept because various diseases may be related to each other. One of these relationships is the relationship between upper and lower levels, for example, diabetes and Class A diabetes, type B diabetes; if there is Class A diabetes, there must be a superior; if a does not exist, it cannot indicate whether the superior exists or not.

1.7 Protocol)

In, protocol is an important knowledge layer concept, indicating the method used for observation. For example, when we take temperature, the thermometer can be placed under the armpit or in the mouth. In some cases, it is necessary to record these different methods. Furthermore, the accuracy and sensitivity of the results can be determined based on different observation methods.

Put this part of information separately in protocol, which is concise and easy to process.

1.8 dual time record)

Observation often has a limited validity period. In the following figure, observation is related to two time records, one indicating the record time and the other indicating the occurrence time, time record can represent both time points and time periods, which can be reflected by its subclass.

1.9 rejected observations

During observation, there are always some objects that are rejected, which may be incorrect, inappropriate, and expired. Then, the observation of this part becomes rejected observation, and used as a subclass of observation. Each rejected observation is related to an observation, which determines that the original observation is a rejected observation.

1.10 active observation, hypothesis (hypothesis) and projection (estimation)

In the medical system, a doctor often needs to make predictions and estimates based on existing observations. These may not be actual measurements. Therefore, the possibility of observation is introduced here, it generates three subclasses: active observation, hypothesis (hypothesis), and projection (estimation. Hypothesis requires more observation to determine. If projection is true, more active observation is required to support it.

The determination of the observed results will attract everyone's attention, but it may be better for doctors to decide this aspect.

1.11 observation Association

In a diagnosis, there are often a series of related observations. In the following model, observation can be associated with observation, in the knowledge layer, the observation concept must also be associated with the observation concept. Such association defines the associative function at the knowledge layer and uses the observation concept as the parameter and the Association result, the rules associated with observations are formulated so that the associated observations can be found for a given observation.

The knowledge layer and the operation layer do not completely form a mirror ing. As you can see, associative function is not a subclass of observation concept. An observation concept may have several associative functions to get it as a result, but a specific observation only has one observation set as evidence.

1.12 observation process

It is necessary to take a look at the observation process. A brief process starts from observation (which may be some measurements, then we can get the relevant observations from the knowledge layer and the model in the previous section, and include these observations in the observations we suggest in the next step. This is a cyclical process.

A more complex rule and description is to find the observation concept from an observation and find the associative function with the observation concept as the parameter, and verify each associative function, the corresponding observation concept can be obtained as the product, and the product can be used as the answer for subsequent observation. You can view the sequence diagram as follows:

A more detailed observation process also includes dividing observation into active observation, hypothesis (hypothesis) and projection (estimation), and projection (estimation) and observe a part of the activities for analysis, subsequent observation, observation of some of the activities based on conflicting observation results to determine whether they can be rejected (rejected observations ).

Conclusion 1.13

From the introduction of the common Quantity Model to the entire observation process, this chapter covers general models applicable to medical systems, in fact, these models are not used in a specific field. If they are similar to your project, you can consider using them or make some changes. In the next chapter, this article introduces that the observation mode is also applicable to enterprise analysis.

Address: http://news.csdn.net/n/20050510/21139.html