Error checking of large land quality databases through data mining based on low frequency associations

• Published in: Land degradation & development 2020-09, Vol.31 (15), p.2146-2155
• Main Authors: Qiu, Xiao‐Qian, Zhu, A‐Xing, Hu, Yue‐Ming, Guo, Yu‐Bin, Shen, Xiao‐Wen
• Format: Article
• Language: English
• Subjects: Cultivated lands; Data mining; error checking; Land degradation; land quality assessment; land quality database; Land use; Low frequencies; low‐frequency data associations; Quality assessment; Quality control
• ISSN: 1085-3278; 1099-145X
• DOI: 10.1002/ldr.3581

The accuracy of databases on land quality, particularly on cultivated land quality, is a prerequisite for land quality assessment and land degradation evaluation. Error checking of land quality databases is an important step in ensuring the accuracy of these land quality databases. The existing methods do not consider the intrinsic relationships among data elements in error checking of land quality databases. This paper explores a new idea for error checking of land quality database through the use of intrinsic relationships that existed in the database. The main assumption behind this idea is that database errors tend to occur at low frequencies and exist as low‐frequency associations with other data items in a database. Thus, these errors can be located by analyzing the combinational relationships between the data items in the database. Based on this idea a new method, low‐frequency data associations (LFDA) through data mining was developed in this paper. The results from control experiments shows that LFDA is effective in locating errors introduced into a land quality database. The applied experiment using the Guangzhou land quality database further confirmed this finding. This research opens a new and significant way for error checking of land quality databases.