M again ================================================================================================== meteorological factory => factor determine the optimal non-linear mapping parameters between ... => determine the optimal non-linear mapping between ... Victor R. et al. Xianlun Tang et al. [16] "theory ." "[8]proposed" areas . capability.Xianlun ... 2. Hybrid Forecasting Method: ARMT-CPSO-BP Neural Network typos Fig.1. The template algorith English is mostly good but sometimes terrible 2.3 (who ) according to the discription of CLS in the 3rd paragraph, the CPSO the authors used is the one proposed by [22] what is the three parametors in the model? what is D-day? in Fig. 2 (should be the day when data were taken) number of input is 7 or The figure is symbolic and more number of imput? We therefore briefly discuss methods helpful for developing parsimonious models. => ? if it does not significantly increase the regression sum of squares sum of squares due to regression. => ?? The stepwise regression procedure combines forward selection with backward elimination. Table 1 should be explained what are X1 ... X7 and what is R? what is ppbv or what is ppb later appeared Table 2 - 4 are not understandable at all. table 5 the figure is ... Final ================================================================================================== Reviewer changed. So most of the below should be pointed out in the first review ... What was pointed out as for BP neural network is now OK. It was corrected. But new reviewer found that the paper is still far from very good for publication. In a scientific article, it is one of the essential rules that author should cleary distinguish what is author's claim from other's claim. As shown below, this paper definitely failed this principle. Mehods are described quite precisely this time, but the gap between the described methods and how the authors apllied them to their own problem is beyond reviewer's imagination. In conclusion, the topic could be a good one if the author further elaborate the paper in the future, but at this stage, the paper is not mature at all. Some explanation is too basic to be in a scientific paper. For example "The mean absolute error (MAE) is the average absolute value of these residual values and the root mean square error (RMSE) is the square root of all squared residuals." This is like an explanation in a textbook. Whilst others need to be explained more specifically. An example is "A careful and systematic approach asks whether..." What exactly is this approach? I'll attach my memo as for expression and as for typo's. Survey is good -------------------------------------------------------------------------------------------------- ARIMA is appeared without explanation ------------------------------------- 2.1 Association rule mining technique ------------------------------------- The subsection 2.1.1 can be understood but should be improved for easier understanding. For example, authors' expression: "The rule x->i_j is satisfied in the set of transactions T with the confidence factor 0<=c<=1 if f at least c% of transactions in T that satisfy X also satisfy I_j." is unclear. what is f in the phrase "if f at least c% ..." or you mean iff? "a rule is satisfied with a confidence c%" is O.K. but what about both "(a set of) t satisfy X" and "(a set of) t satisfy I_j"? Both are not difiend. The verb "bought" in "if it bought the item" is not appropriate as a theory description, easy example to understand though. "(may be considered later)" is not appropriate. If actually considered later, it would be "(See later more in detail.) of something like that. "To do this task efficiently, we use some estimation tools and some pruning techniques." should be more specifically. That is, "some" makes the explanation unclear. The algorithm shown in Fig. 1 is authors' original or a proposal by others (maybe by [18])? In what author wrote "The connection weights are assigned initial values first," what are initial values? A random value? Authors wrote "The error between the predicted and actual output values is back-propagated via the network for updating the weights [19]." But the original idea is not by [19]. The same holds [20] and [21]. These papers might be just papers in which author learned what the BP is. Author wrote "Theoretically, neural networks can simulate any kind of data pattern given sufficient training." Who claim this? Not authors' theory right? Then reference should be shown. No description about W_{ij} which is important because connection from i to j or opposit is crucial. Authors notation is different than standard usage. -------------------------------------------------- 2.3. Chaotic particle Swarm Optimization Algorithm -------------------------------------------------- Who proposed PSO for the first time? It should be added in the Reference. Otherwisze it looks like authors propose. The variable g in (9) is not explisitly explained Also in (9) and others we read "where k = C, epsilon, sigma" but what is this? Describe what happens when V_ki becomes the range The description "until the pre-defined termination criterion is satisfied" is not the idea only in [17], but quite general criterion. So delete [17]. Is [15] the original proposal of CPSO? It is not, isn't it? ------------------------- 2.3.2. Chaotic PSO Method ------------------------- The first pragraph of 2.3.2 should be refiend. What is l or what should it be called in (11)? What is globalbesti? Delete [24] in "The BP algorithm for multi-layer neural network [24]," for the reason already mentioned. ------------------------ 2.4. Performance Indices ------------------------ ------------------------------- 3.3 Instruments and Measurement ------------------------------- what is ML9800, ML9810B etc? ------------------------------ 4.1 Independent Input Variable ------------------------------ "None of the inference methods described above performs reliably if factors are missing from the model." might be bette to be "None of the inference methods described above performs reliably if any of onf the input factors are missing from the model." "We therefore briefly discuss methods helpful for developing parsimonious models." might be better to be "We therefore briefly discuss methods which help us develope a parsimonious model." What does it mean by "the regression sum of squares sum of squares due to regression"? Also "the regression sum of squares sum of squares due to regression" is difficult to understand. Authors wrote "Firstly, we prepare seven factors for selection." Then what comes secondly? Or, why these seven factors are chosen firstly? What is R in Table 1? what is ppbv? what is S1(i) and T1(i) in table 2? what is sigh>? The similar umbiguity exists also in table 2 Table 3 is not understandable. Thse three tables are more like an author's memo for the raw data to prepare for their paper. eng 0.00mm Caption of Fig.4. should be "Comparisons of three models through the prediction of the ozone levels (1-24 July 2009) using the testing data set (25-30 July 2009)." or "Comparisons of three models through the prediction of the ozone levels using the testing data set." ------------------------- 4.2 Result and discussion ------------------------- Title of this subsection should be "Results and discussion" What actually means by "A range of evaluative statistics"? Reviewer understand what it means, but as an expression it is not good at all. Eng, typo -------------------------------------------------------------------------------------------------- => "should be," "might be," "is better to be," or comment follows after this symbol punctuation equation all equations if not in the text should be end with period. wellbeing => well being urban areas . out performs => outperforms trial approaches => trial and error approaches An improved PSO ... => better to be particle swarm optimization because it is firstly appeared capability.Xianlun chaotic swarm optimization algorithm (CPSO) => chaotic particle swarm optimization algorithm (CPSO) meteorological factory => factor I_x appearing in the consequent => I_x in the consequent I_y appearing in the antecedent => I_x in the antecedent X appearing in the consequent => X in the consequent Y appearing in the antecedent => Y in the antecedent possible- => possible predefine itemset => predefined itemset ,k>2 at the most k => at most k minsuppori (in Fig.1)=> minsupport algorith => algorithm set[18]. => set [18]. network.The => network. The Steps in training BP network as followed => Steps in training BP network are as follows Select training pair from the training set and apply the input vector to the inputlayer. => Select training pair from the training set and apply them one by one as the input vector to the input layer. Each particle flies => Each swarm particle flies can be founded => can be found in respect with => with respect to electiveness => effectiveness? in noon => at noon 13:00 in the noon => 13:00 (in the afternoon) 0.00mm => 0.00 mm about 30 days => for 30 days "there have 48 times measures" => ? "This method is developed based on association rule mining technique, chaotic particle swarm optimization algorithm (CPSO) and BP neural network." => Be consistent. Sciences,2008 not-yet below 2.1.2 Discovering large itemsets