POI POI data validation is to ensure the correctness of the data, and the results for the correctness check data released last guarantee, so the data need to check in more latitude, and strive correctness POI.
Change data is divided into state change, change key fields and details of field changes. The largest state in which the impact of changes, can cause changes in the offline state online POI, and the details of the underlying field and the field will affect the validity of the business. Notice the change causes a variety of data, which can be divided into three categories:
1. Data source changes
2. The processing policy changes
3. systematic errors
1. The data source changes, the new data, offline access large amounts of data source or change existing sources could cause POI data changes;
2. The processing policy change process for policy changes in state or key fields can cause data changes, such as adding a filtering policy name, if the name is not legitimate it will filter the data directly off the assembly line, if this strategy will result in a greater impact surface a large area of data offline; and if the name of the policy will not only legitimate character mask then the data at the line, while the name may differ from the actual POI.
3. refers to systematic errors in the data processing bug causing errors, data errors that affect the data side effect would be more concentrated due will be relatively large, but also the easiest to find, just compare the new data with the results whether on a version of the data change, whether to change the rate of significant changes in data can be an early warning of systemic errors.
Data validation:
Field validation: Value Type field POI data, range check, implementation rules can be written into the configuration, the separation can be achieved when the code rules by modifying the configuration changes
Rule checking: the whole set of data validation rules, such as a field must have a value well and have a field validation rules, the proportion of the value of a field there, and when it does not conform to the rules intercept
Change verification: Compare this release and change data released last change data, including data and new off the assembly line, and a data field issued two coexist, and when there is no change of name, address, coordinates as key fields its rate of change should be less than 1 ‰, if you change the value of the surplus should confirm the data source, whether the policy has changed, and the change results in line with expectations.
Thus policy data checksum, data access should linkage; set change POI data access when the data changes over time as the access data may cause changes; policy change due to the data and the expected result of a change should result have informed the process and cross-checked with the results of the check; when the expected change with time does not match the pre-designed, should be introduced artificial evaluation, the evaluation of this data change if there is a positive impact on the data to determine whether the line.