Machine learning aids to solve a problem intelligently after gathering data. A lot of data is required by the algorithm to learn and find a solution. Algorithms define a series of steps to solve a problem. Algorithms can be pre-defined after learning some data or can continuously evolve after gathering more data to provide a better solution. Due to the advancement in technology, there is a huge amount of data which is collected and can be analysed to define an algorithm. For example, a supermarket stores the related information of the products brought by a customer, frequent products which are brought together and manage the available inventory. This data can help to extract patterns regarding buying patterns. A machine learning algorithm can be designed and implemented to analyse a customer’s needs and how to arrange the products smartly in the supermarket. Machine learning is divided in few techniques in the book (Mohammed et al. 2016) as mentioned below:
- Supervised Learning:
The data is pre labelled which is used for training an algorithm. Unlabelled is used for testing the model. The algorithm is designed on the basis of training data which is known and used to find the results for unlabelled testing data. The manual labelling process can contain errors.
- Semi-Supervised learning:
Semi-Supervised learning contains most of the labeled data and some unlabelled data for training. It reduces the cost by keeping unlabeled data as well and also maintains the accuracy of predictions by keeping labeled data.
- Unsupervised Learning:
The data does not contain any labels and are unknown before. Algorithms group and create clusters of data according to the properties and similarities. Clustering techniques are used in unsupervised learning.
- Reinforcement Learning:
Reinforcement Learning contains no data rather it considers the observations from the environment to design a model for predictions.
Applications of Machine learning algorithms in Agriculture:
With the advancement in technology, a lot of sectors make use of it to increase profits. One such example is, Agriculture. Smart Agriculture makes use of sensors and machinery to gather information about the crops and provide better results. Several operations in agriculture use IoT, Big Data and Cloud to analyse, evaluate and control the crops (3M 2018). It helps to utilize the resources effectively and increase the productivity of the crops. Sensors in the fields determine the best time to harvest, create reports and gather data to monitor the health of the crops.
There are different applications of Machine learning algorithms in Smart Agriculture. One such example is mentioned in (Jiang et al. 2020) to distinguish weeds and crops. A framework is designed using Graph convolutional network (GAN) and applied on images to classify their labels. The features of an image were analysed and sent to the classifier to identify the labels. Models were created based on the labeled data and updated according to the features of unlabelled images. The accuracy was improved using a semi-supervised approach, that is, the training data set used consists of labeled and unlabeled images.
Moreover, (Zhou et al. 2017) describes an implementation of the DBSCAN algorithm. It uses clustering to identify the labels of the data points. Firstly, a core point is identified on the basis of the most points closed to a certain point. Secondly, a boundary is identified to distinguish the data points according to the distance and link from the core point. Then, outlier data points are identified which help to analyse the reasons. The irregular shaped clusters helped to predict the reasons for low productivity. For example, some weather conditions might have affected the production of the crops and resulted in outlier data. It can assist the owners to plan and find solutions to increase productivity during those times. IoT enables us to gather a lot of real-time data through sensors in the field, hence, a lot of analysis can be done. It can calculate the right amount of water for the crops, what should be the amount of pesticides, and which crops should be grown beneficially according to the weather. A lot of methodologies can be used to achieve improvements in agriculture.
Another implementation of a clustering algorithm in agriculture is explained in (Machica, Gerardo, and Medina 2019). IoT helps to identify the anomalies in the data gathered from the crops. The anomalies can help the farmers to identify if there are any problems and alert the farmers to take necessary action. The paper describes a use of a superimposed classification algorithm to detect any anomalies in the crops. It works to find any anomalies by gathering data of crops in different locations. It uses data mining techniques to clean, understand, and prepare models.
Similarly, there are other applications that use clustering algorithms to identify the crop. One implementation is mentioned in (Zhang and Chau 2009) which uses KNN classifier to predict the class of the plants with the help of leaf images. Firstly, it selects a few features which can be used to classify the images. Then, a euclidean distance metric is used to classify the data points. The greater distance between them represents less similarity, whereas, less distance indicates similar data points. The proposed approach observed the results on k binary classification problems. If there will be multiple classes, it will be divided into binary classification problems and then applied to the proposed solution. Also, it used a 10-fold cross validation technique to get better results. A k-fold cross validation technique divides the data, k times, into training and testing datasets and evaluates the results. The algorithm uses the train-test split with most accurate results.
In conclusion, there are a lot of different implementations of algorithms in agriculture, which improves productivity and aids the farmers to monitor the fields. New and improved technologies can be used to reduce the errors and increase the growth of the crops. It can provide great advantages to agriculture.
References:
3M. Smart Farming: Data Enabling the Future of Agriculture, 2018. https://www.youtube.com/watch?v=LaMvMgdJC58.
Jiang, Honghua, Chuanyin Zhang, Yongliang Qiao, Zhao Zhang, Wenjing Zhang, and Changqing Song. “CNN Feature Based Graph Convolutional Network for Weed and Crop Recognition in Smart Farming.” Computers and Electronics in Agriculture 174 (2020): 105450
Machica, Ivy Kim D., Bobby D. Gerardo, and Ruji P. Medina. “Superimposed Rule-Based Classification Algorithm in IoT,” 2019.
Mohammed, Mohssen, Muhammad Badruddin Khan, and Eihab Bashier Mohammed Bashier. Machine learning: algorithms and applications. Crc Press, 2016.
Zhang, Shanwen, and Kwok-Wing Chau. “Dimension Reduction Using Semi-Supervised Locally Linear Embedding for Plant Leaf Classification.” In International Conference on Intelligent Computing, 948–55. Springer, 2009.
Zhou, Yifan, Wei Hu, Yong Min, Le Zheng, Baisi Liu, Rui Yu, and Yu Dong. “A Semi-Supervised Anomaly Detection Method for Wind Farm Power Data Preprocessing.” In 2017 IEEE Power & Energy Society General Meeting, 1–5. IEEE, 2017.
Sr. Software Engineer
